Wireless sensor networks for condition monitoring in the railway industry a survey

1088 IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS, VOL. 16, NO. 3, JUNE 2015
Wireless Sensor Networks for Condition Monitoring
in the Railway Industry: A Survey
Victoria J. Hodge, Simon O’Keefe, Michael Weeks, and Anthony Moulds
Abstract—In recent years, the range of sensing technologies has
expanded rapidly, whereas sensor devices have become cheaper.
This has led to a rapid expansion in condition monitoring of
systems, structures, vehicles, and machinery using sensors. Key
factors are the recent advances in networking technologies such
as wireless communication and mobile ad hoc networking coupled
with the technology to integrate devices. Wireless sensor networks
(WSNs) can be used for monitoring the railway infrastructure
such as bridges, rail tracks, track beds, and track equipment along
with vehicle health monitoring such as chassis, bogies, wheels,
and wagons. Condition monitoring reduces human inspection re-
quirements through automated monitoring, reduces maintenance
through detecting faults before they escalate, and improves safety
and reliability. This is vital for the development, upgrading, and
expansion of railway networks. This paper surveys these wireless
sensors network technology for monitoring in the railway indus-
try for analyzing systems, structures, vehicles, and machinery.
This paper focuses on practical engineering solutions, principally,
which sensor devices are used and what they are used for; and
the identification of sensor configurations and network topologies.
It identifies their respective motivations and distinguishes their
advantages and disadvantages in a comparative review.
Index Terms—Asset management, condition monitoring, deci-
sion support systems, event detection, maintenance engineering,
preventive maintenance, railway engineering, wireless sensor net-
works (WSNs).
I. INTRODUCTION
EXPERTS estimate that the railway industry will receive
US$300 billion worth of global investment for devel-
opment, upgrading, and expansion over the five years from
2009 [43]. Ollier [98] noted that effective management of rail
infrastructure will be vital to this development, upgrading, and
expansion, particularly if coupled with a move to intelligent
infrastructure [39]. A key part of the management will be con-
dition monitoring. Condition monitoring detects and identifies
deterioration in structures and infrastructure before the deterio-
ration causes a failure or prevents rail operations. In simple con-
dition monitoring, sensors monitor the condition of a structure
or machinery. If the sensor readings reach a predetermined limit
or fault condition, then an alarm is activated. However, this sim-
Manuscript received April 2, 2014; revised August 8, 2014; accepted
October 17, 2014. Date of publication November 20, 2014; date of current
version May 29, 2015. This work was supported by the U.K. Engineering and
Physical Sciences Research Council under Grant EP/J012343/1. The Associate
Editor for this paper was X. Cheng.
The authors are with the Department of Computer Science, University of
York, YO10 5GH York, U.K. (e-mail: victoria.hodge@york.ac.uk; simon.
okeefe@york.ac.uk; michael.weeks@york.ac.uk; anthony.moulds@york.
ac.uk).
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TITS.2014.2366512
Fig. 1. Figure shows a typical WSN setup for railway condition monitoring.
Sensor devices are mounted on boards attached to the object being monitored;
examples include track, bridges, or train mechanics. One or more sensors are
mounted on a sensor board (node) (see also Fig. 2). The sensor nodes communi-
cate with the base station using a wireless transmission protocol; examples in-
clude Bluetooth and Wi-Fi. The base station collates data and transmits it to the
control center server possibly through satellite or GPRS. There are variations
on this setup. In some systems, the sensor nodes may communicate directly
with the server rather than via the base station. In other systems, the user ac-
cesses the data directly via the base station.
plistic approach may lead to a large number of false alarms and
missed failures [36]. It only provides local analysis but does not
take advantage of the superior capabilities when the sensors are
networked and their data processed collectively. Integrated data
processing allows an overall picture of an asset’s condition to be
achieved and overall condition trends to be determined [97].
In recent years, networking technologies such as wireless
communication and mobile ad hoc networking coupled with
the technology to integrate devices have rapidly developed.
The new technologies allow vast numbers of distributed sen-
sors to be networked [5], [6], [37], [45], [122] to constantly
monitor machines, systems, and environments. Wireless sensor
networks (WSNs) [5], [134] are wireless networks of spatially
distributed and autonomous devices. They use sensors to co-
operatively monitor infrastructure, structures, and machinery.
A typical WSN for railway applications is shown in Fig. 1.
Each sensor node generally has a radio transceiver, a small
microcontroller, and an energy source, usually a battery (see
Section II-C for more detail). WSNs and data analytics allow
the railways to turn data into intelligence [43]. They provide
decision support through continuous real-time data capture
and analysis to identify faults [52]. The data from distributed
systems such as sensor networks are constantly monitored using
classification [56], [57], prediction [85], or anomaly detection
[61] to determine the current and future status of the distributed
network. Lopez-Higuera et al. [78] developed a staircase of
structural health monitoring, where the higher the level, the
This work is licensed under a Creative Commons Attribution 3.0 License. For more information, see http://creativecommons.org/licenses/by/3.0/
www.redpel.com+917620593389

HODGE et al.: WSNS FOR CONDITION MONITORING IN THE RAILWAY INDUSTRY: A SURVEY 1089
higher the complexity and functionality. The simplest Level
1 systems detect the presence of damage without locating it,
whereas Level 2 provides location information. A Level 3
system is able to grade the degree of damage and a Level 4
system can estimate the consequences of the damage and
remaining service life. Finally, a Level 5 system will comprise
complex hardware, custom algorithms and software to allow
the diagnosis and/or the prognosis and even to recommend the
solution to a problem.
WSN monitoring provides continuous and near real-time
data acquisition and autonomous data acquisition (no supervi-
sion is required); increased frequency of monitoring compared
with manual inspection; improved data accessibility, data man-
agement, and data use compared with non-networked systems
as all data can be collected and processed centrally; the ability
to combine data from a wide variety of sensors; intelligent
analysis of data to “predict and prevent” events using intelligent
algorithms; the ability to turn data into information about the
status of important structures, infrastructure and machinery;
and, a global data view that allows trending information to
be determined where degradation is happening slowly over a
relatively long period of time.
WSN monitoring can be used to:
1) maintain process tolerances;
2) verify and protect machine, systems and process stability;
3) detect maintenance requirements;
4) minimize downtime;
5) prevent failures and save businesses money and time;
6) request maintenance based on the prediction of failure
rather than maintenance running to a standard schedule
or being requested following an actual failure.
There are a number of challenges with WSNs. They generate
large amounts of data at rapid rates and often on an ad hoc basis.
Data may be produced from multisources that have to be fused.
The systems and structures monitored using sensors often ex-
hibit complex behavior, which is difficult to understand and
interpret [102]. Hence, the data must be carefully managed to
provide a view of the system status. Sensor data are very noisy
and sensors themselves can become defective wherever they are
installed. Sensor data may contain errors, particularly where
the sensors are subject to harsh conditions as this exacerbates
sensor and communication failures. Sensor networks often have
to be installed in challenging environments to be able to monitor
structures and infrastructure. For example, Palo [99] noted that
their system had to work in extreme conditions in Sweden, with
a temperature range between +25◦
C and −40◦
C and with large
quantities of snow. Grudén et al. [55] mounted sensors on the
train’s bogies to monitor bogie temperatures and noted that train
environments are very harsh environments for electronics with
high accelerations and large shocks. The sensors need to be
carefully located to ensure their measurements are useful and
do not replicate the measurements of other sensors, which can
skew the distribution of the collected data. The type of sensor
used needs to be carefully considered to ensure the maximum
value and the best quality data. WSNs can use a set of homoge-
neous or heterogeneous sensors. Sensors are often located away
from energy supplies, thus require either batteries or some form
of local energy generation to power them. If there are errors
in transmission across the WSN, then data may be missing.
These last two points form a paradox, WSNs need to minimize
energy usage yet communication needs to be maximally effi-
cient and communication requires energy.
This survey paper describes WSNs for railway condition
monitoring focusing on systems described in the academic
literature. In this survey, “sensor” refers to an individual device
such as an accelerometer or strain gauge. “Node” refers to a
whole sensor unit that comprises the sensor, a power supply,
a data transmitter/receiver, and a microcontroller. Section II of
the survey discusses the design and range of sensor devices,
Section III analyzes the network topology and transmission
techniques, and Section IV reviews the sensors used in the lit-
erature to monitor railway equipment and structures. Section V
draws conclusions, and Section VI examines future work in the
area of railway monitoring.
II. SENSOR DESIGN
There are a multitude of sensors types used in railway con-
dition monitoring for analyzing different aspects of structures,
infrastructure, and machinery (see Table I in the Appendix).
The devices in Table I are transducers. Transducers convert
energy from one form to another. In the case of sensors, the
device typically converts a measured mechanical signal into an
electrical signal. Most railway sensors fall under the umbrella
type microelectromechanical systems (MEMS). MEMS are
small, integrated devices, or systems that combine electrical
and mechanical components. They are cheap and efficient [70].
Sensor design requires a tradeoff between functionality and
power consumption, with functionality often coming at the
cost of power. Condition monitoring systems for railways are
often deployed in remote or inaccessible locations, where there
is no wired power supply available. Hence, the sensors must
receive power from either batteries or local energy generation.
MEMS sensors have the advantage that they can be produced
to consume ultralow power [19].
A. Measurement Sensors
Table I provides an overview of the sensor types used in
railway condition monitoring and the measurements produced.
Sensor devices may also return additional ambient mea-
surements such as temperature and humidity, and sensor data
can be combined with the measurements from other systems.
Rabatel et al. [103] noted that the behavior of assets is often
affected by external factors. These external factors are encap-
sulated by contextual data, which describe the ambient running
conditions. It is important to take them into account and build
models for condition monitoring that consider these ambient
conditions. Hence, Elia et al. [44], Rabatel et al. [103], and
others incorporated wider ranging contextual data with the
sensor data. Elia et al. [44] included train speed, position along
the line (longitude and latitude from GPS system), and wind
speed and direction when analyzing bogie vibrations, whereas
Rabatel et al. [103] included structural criteria (route) and
environmental criteria (weather, travel characteristics, travel
duration) when analyzing bogie temperatures.

Fig. 2. Diagram of the composition of a typical sensor node. Multiple sensors
may be mounted on the node. The microcontrollers used in sensor nodes are
ultralow-power microcontrollers to conserve energy.
B. Sensor Nodes
Sensor devices are mounted on boards. The boards form
a platform combining mobile computing and wireless com-
munication (media access control (MAC), routing and error
detection) with sensor devices as shown in Fig. 2. Many sen-
sor nodes are moving toward enabling data interrogation and
autonomous data processing to identify anomalies in situ.
Hence, many data processing algorithms have also been embed-
ded in wireless sensor nodes for autonomous execution [80].
The boards generally comprise one or more wireless sensors,
a microcontroller, transceiver, data storage (memory), and a
power source. If the data sampling rate is low, for example,
when monitoring the growth of a crack over a period of time,
then the microcontroller, memory, and transceiver power can
all be low. However, for frequent sampling such as monitoring
the temperature of a mechanical bearing every second, then
the microcontroller, memory, and transceiver must be more
powerful. Lynch and Loh [80] provided a thorough overview
of sensor boards for WSNs.
C. Sensor Power
Sensor nodes often use batteries as their energy source.
Batteries have a finite life. The batteries in an accelerome-
ter monitoring a train pantograph only lasted 20 days [97].
Chebrolu et al. [32] were able to extend battery life to
1.5 years for 4 AA batteries by only waking their sensors when
required and sleeping otherwise. However, even with extended
life, fitting and replacing batteries is not always possible in
remote and inaccessible railway locations. The nodes may also
be embedded in the monitored structure or mechanics of a train,
where no access to batteries or wired connection is possible.
Hence, authors have considered alternative energy sources for
nodes. These need to be reliable to allow the WSN to function
for long periods of time with no human involvement and able
to generate sufficient power.
Nagode et al. [89], Nelson et al. [92], Nuffer and Bein
[96], Pourghodrat [101], Tan et al. [120], and Wischke et al.
[130] investigated ambient energy harvesting technology for
powering WSNs. Piezoelectric materials can produce electricity
as a response to mechanical strain such as the strain exerted on
the track or sleepers by a passing train. Piezoelectric materials
are robust, able to withstand harsh environmental conditions,
and can deliver a large amount of energy [55]. Nelson et al.
[92], Nuffer and Bein [96], Pourghodrat [101], Tan et al. [120],
and Wischke et al. [130] investigated piezoelectric vibration
harvesters that derive energy from the vertical displacement or
vibration of tracks, ties, and sleepers. Wischke et al. [130] were
able to demonstrate that sufficient energy could be harvested to
supply a microcontroller with a radio-frequency (RF) interface.
Tan et al. [120] noted that their rate of energy harvesting
(≈1 mW) was lower than the typical rate of energy consump-
tion of sensor systems (≈10–100 mW). Hence, they stored
the harvested energy using supercapacitors and the sensors
only drew energy when sufficient stored energy was available.
Nagode et al. [89], Nelson et al. [92], and Tan et al. [120] all
used inductive coils for harvesting. Electromagnetic harvesters
such as induction coils have a higher power density than piezo-
electric [69]. Nelson et al. and Tan et al. used the vertical move-
ment of the track to generate energy. Nagode et al. designed
and built two prototypes of a vibration-based electromagnetic
(inductive coil) system. They both harvested mechanical energy
from the movement of a rail wagon’s suspension springs to
power sensors mounted on the wagons. The first prototype
used the movement of the springs to generate linear motion in
magnets to produce a voltage. The second prototype converted
the linear motion of the wagon’s suspension coils into rotary
motion then magnified and rectified the motion to turn a gener-
ator. Pourghodrat [101] developed a hydraulic energy harvester,
which generated energy from compression and relaxation of
fluid driving a hydraulic motor. It used the vertical displacement
of the track for energy generation. Hydraulic harvesters can
generate energy from smaller track deflections compared with
electromagnetic, but the devices are still in development [101].
A number of authors have investigated using solar energy
harvesting to power sensors. Bischoff et al. [19] investigated a
solar energy rechargeable battery-powered base station. Sekula
and Kolakowski [111] powered their piezoelectric strain sen-
sors using accumulators permanently charged by photovoltaic
modules. Solar power is a proven way to generate energy.
However, it has a number of practical drawbacks for WSNs.
Solar panels used to generate energy need frequent cleaning
for optimal performance [55]. Sunlight varies on a daily basis,
and sunlight is very limited in winter months in many global
regions and power generation is adversely affected by bad
weather. Hence, Wolfs et al. [131] powered accelerometers
and gyroscopes using solar power with battery backup and
Waki [125] powered a temperature sensor by a solar cell with a
double-layer capacitor power supply for back up. These double
layer capacitors have a longer life than a battery.
If the sensor nodes can be put into “sleep” mode and only
woken when they are required, then their energy usage can be
minimized. A number of bridge monitoring systems use this
wake-on-event two-layer event detection. It requires two types
of sensors: one set have lower power consumption to perform
event detection and wake the second set of sensors using inter-
rupts, [19], [32], [46], [70], [73] and [107]. Bischoff et al. [19],
Feltrin [46] and Kruger et al. [70] detected approaching trains
using ultralow-power MEMS acceleration sensors. These mea-
sured the vibrations of the bridge and detected the increased vi-
brations when a train approached. Similarly, Ledeczi et al. [73]

used standard strain gauges to sense the approaching train and
woke the system by detecting changes in stress on the bridge.
Sala et al. [107] used activation sensors type not specified) to
wake their system for monitoring steel truss bridges. In con-
trast, Townsend and Arms [123] used a similar technique but
with only one set of track-mounted strain gauges, which varied
their sampling rate depending whether a train is present. They
operated in low sampling mode of approximately 6 Hz to detect
the increase in strain when a train approached. Once detected,
the sensors increased their sampling rate and generated a strain
waveform for data analysis. The authors stated that this reduced
the power from 30 mA for continuous mode to less than 1 mA
for event detection. Chen et al. [35] introduced a prototype
system that developed this further by detecting approaching
trains, waking the measuring sensors on the bridge trusses and
then downloading the measured data onto the passing train,
which acts as a data mule. Another data muling approach is the
BriMon system [32], where trains were fitted with a beacon. A
bridge-mounted high-gain antenna detected the train’s beacon
30+ seconds before the train approached and woke the battery-
operated accelerometers on the bridge. The accelerometers
measured the vibrations of the passing train on the bridge and
uploaded their data to the passing train. This data muling only
requires local data transmission, which reduces the power usage
further.
III. NETWORK DESIGN
WSNs enable continuous real-time capture of data. However,
WSNs need to be able to handle the harshness of outdoor long-
term condition monitoring; often in hostile environments and
must minimize energy usage as the nodes are not attached to
a wired power supply. They typically use low-power sensors
powered by batteries although authors are investigating alterna-
tive power supplies such as local energy generation. Hence, the
network to enable data capture has to be carefully designed to
overcome these factors and prevent transmission errors, latency,
network outages, missing data, or corrupted data.
A. Base Station
The base station controls the sensor nodes and acts as a gate-
way for data transmission to a remote server. The sensor nodes
use short-range communication such as Wi-Fi or Bluetooth to
transmit data to the base station. The base station uses long-
range communication such as GPRS or satellite to transmit
collated data back to a server at a control center. It has a more
powerful processor and more memory than the sensor nodes
to allow it to collate data from multiple sensors. Hence, the
base station requires more power. For example, train monitoring
systems have battery-powered sensor nodes in the carriages but
place the base station in the locomotive, where more power is
available from the train engine.
B. Relay Nodes
Sensor nodes are energy-constrained, thus only having a
short transmission range. If they cannot reach the base station,
then higher capability relay nodes may be used. This is a
particular issue in restricted environments such as railway tun-
nels, along the length of trains or along railway tracks. Relays
ensure connectivity by relaying the data signal from the sensor
node to the base station via one or more relay nodes using
wireless multihop paths. Network nodes can be multifunctional.
Hoult et al. [62] and Shafiullah et al. [112] both used sensor
nodes that also function as relay nodes to transmit data.
C. Network Model
WSNs for railway applications follow the open systems
interconnection model [137]. A typical WSN has five layers
in a protocol stack with three planes [5], [6] to transmit data
from the sensor nodes to the base station. The five layers
from lowest to highest are the physical layer defines how the
sensors transmit their data to the network. The data link layer
specifies the network topology and connects nodes to each
other. It uses MAC to control data access and permissions and a
logical link control to control error checking and data packet
synchronization. This is a key layer in railway monitoring
WSNs as they may be monitoring difficult environments such
as tunnels or monitoring moving trains. The topology must be
carefully designed. The network layer routes the data through
the network as packets. Railway WSNs are energy constrained;
thus, routing has to be carefully designed to work within the
available energy supply. The transport layer controls the send-
ing and receiving of data. Finally, the application layer allows
application software to access the data.
The size of the transmission varies from bits in the lowest
(physical) layer, bytes, and frames in the data link layer,
packets in the network layer, segments in the transport layer,
and data in the highest (application) layer.
The power management plane manages the power con-
sumption of sensor nodes. The mobility management plane
identifies and records the movement of sensor nodes to ensure
that data can be routed from the node to the control center
and so that sensor nodes are aware of their neighbors for com-
munication. This is important for WSNs monitoring moving
trains. The task management plane balances and coordinates
network regions (groups of nodes) and ensures that they gene-
rate data according to their power level.
D. Sensor Network Topology (Data Linkage)
The topology is constrained by the requirements of the moni-
toring and by the physical environment. Sensor nodes can be ar-
ranged in either an ad hoc or a preplanned configuration [134].
In ad hoc arrangement, sensor nodes are randomly placed in the
monitored area. In preplanned arrangement, sensor nodes are
arranged in either a grid, optimal placement [100] (described
next), 2-D or 3-D placement [134].
Determining the optimum node placement is a complex
task and often requires a tradeoff. The network configuration
can be optimized against a number of different constraints.
A network may minimize relay nodes, may need to ensure a
minimum level of service (include a certain level of redun-
dancy), minimize energy usage to preserve battery life, or may

Fig. 3. Diagram of a star, tree, and mesh network topology (from left to right)
where the circles represent sensor nodes and the boxes are base stations.
Fig. 4. In a typical WSN, the sensor nodes are arranged in a star, tree, or
mesh topology and transmit data to the base station, which relays the data to
the control center via longer range communications.
need to ensure accessibility of the nodes [116]. For example,
Pascale et al. [100] optimized a WSN for signaling and control
in railway stations to calculate the minimum number of nodes
required. The authors had to optimize the tradeoff between re-
liability versus communication delay. Determining the best set-
tings proved difﬁcult as it required iterative graph optimization,
which is computationally complex. They split the optimization
into two tasks: deployment optimization to minimize the num-
ber of nodes while maintaining a minimum level of reliabil-
ity, and scheduling optimization which allocates nodes while
maintaining a maximum level of delay. The authors concede
that the reliability versus communication delay tradeoff cannot
be solved for all situations. The railway environment can also
place constraints on the network topology. Signal propagation
within a rail tunnel is different from propagation outside [77],
[88], [116]. The complex geometry and rough surfaces of rail
tunnels degrade RF signals. Conversely, the regular, tube-like
shape can act as an “oversized waveguide” [88]. Therefore,
Stajano et al. [116] and Liu et al. [77] developed models to
predict the signal strength at each node in a tunnel. Node
placement can then be optimized using an algorithm based
on parameters such as transmission strength, battery life, and
redundancy between the sensor nodes and the base station.
The natural topology choices available for WSNs in rail ap-
plications are star, tree, and mesh topologies (see Figs. 3 and 4).
In a star topology (single point-to-multipoint), there is a single
base station that can send and/or receive messages to/from a
number of remote sensor nodes. The remote nodes can only
communicate directly with the base station but not with each
other. This topology is simple; the point-to-point connection
will result in fewer transmissions and fewer collisions; and, it
has low communication latency. The disadvantages are the base
station must be within transmission range of all nodes, remote
nodes must increase their transmission power to overcome
signal attenuation and loss, and it is not robust due to its
dependence on a single base station to manage the network.
A tree topology is a hierarchy of nodes, with the root node
serving client nodes that, in turn, serve other lower level nodes.
In a tree, the nodes can be grouped at each level. The network is
scalable. Messages pass from the sensor nodes through the tree
branches to the root. Sensor nodes can communicate with their
parent or with other nodes in a group within transmission range.
In contrast, in a mesh topology, any node in the network can
communicate with any other node in the network that is within
transmission range. This enables relaying of messages (mul-
tihop communication) keeping the transmission power low,
saving energy compared with transmitting directly from sensor
nodes to base station in a single hop, and allowing nodes to
communicate indirectly with nodes outside their transmission
range. The network can be easily extended by adding new nodes
and also incorporates fault tolerance as failed nodes can be
bypassed using alternative routes.
E. Communications Medium
There are many communication techniques used in WSNs
in railways [3], [108]. Intersensor communications and sensor
to base station transmission are usually short range. The base
station transmits gathered data back to the control center,
and this requires long-range communication. WSNs can use
technologies based on standard mobile telephony (Bluetooth,
GSM, GPRS, and UMTS) or broadband techniques such as
Wi-Fi (IEEE 802.11, 2007), wireless personal area networks
(WPANs) (IEEE 802.15.4 and ZigBee [11]) or WiMax [2]
(IEEE 802.16.2, 2004). ZigBee enhances IEEE 802.15.4 by
including authentication, data encryption for security, and data
routing and forwarding. Mobile telephony and WiFi are peer-
to-peer only, whereas WPANs and WiMax can connect many
devices in a mesh topology, where any node can communicate
with any other node. In addition, Wi-Fi, WPANs and WiMax
allow communication at much higher speeds and bandwidths
than mobile telephony and are cheaper to set up. However,
mobile telephony has better coverage and range than Wi-Fi and
WPANs. WiMax can operate over long distances. However,
there is a tradeoff: either high bitrates or long distances but
not both. An important recent communication advance has
been GSM-R (Global System for Mobile Communications-
Railway), which is an adaptation of GSM telephony for railway
applications. It is designed for information exchange between
trains and control centers. Its main advantages are its low cost
and worldwide availability. Another common technique for
railways is the radio system Terrestrial Trunked Radio, which is
a standard for data communication for closed user groups [108].
It is a private mobile telephone system and is easily deployed
using a series of antennas at stations or control centers along
the railway routes.
F. Transmission and Routing
Sensor nodes use the greatest amount of energy during data
communication. The transceiver in the sensor node contains
frequency synthesizers, mixers, voltage control oscillators,

phase-locked loops, and power amplifiers [6]. These all con-
sume power. Thus, data transmission and routing has to be
carefully designed to minimize the energy usage.
When messages are transmitted through tree or mesh net-
works, they have to be routed to ensure they pass from source
to destination (sensor node to base station). Message latency
increases as the number of hops required to transmit the mes-
sage increases; thus, the routing protocol must be carefully
designed for the particular WSN topology. A number of routing
protocols have been proposed recently (see [4]). If the base
station fails in any topology, then the routing also fails. Thus,
when Mahlknecht and Madani [81] used a daisy chain (tree)
topology, they ensured that the nodes in the chain could also
directly communicate with the global communication network
if no base station was accessible. Another important consid-
eration of multihop communication is interference prevention.
For train-mounted WSNs, if two trains, Train A and Train B,
are within radio communications range of each other, then data
from sensors on Train A must never be transmitted to Train B
and vice versa [104]. Reason et al. [104] used identifiers to bind
the sensors to the individual train’s network.
The sensor nodes need to transmit data reliably via the
network despite having only a limited power supply avail-
able. Hence, energy efficiency and reliability are both vital
for WSNs. One important disadvantage of multihop (tree or
mesh), in general, is that the power consumption of the nodes
near the base station is much higher as they are used in data
transmission to the base station and hence suffer energy drain.
A possible solution to minimize energy usage and maximize
fault tolerance is to use multihoming. This is used in the Internet
and provides multiple connections (multiple routes) between
points thus introducing a degree of fault tolerance and ensuring
that no extra energy is required to transmit data. Aboelela et al.
[1] mapped every node in the network to two clusters. One set
of nodes operated as guardians of the clusters and cluster nodes
transmitted via two designated cluster guardians. Each cluster
guardian thus acted as a relay node and forwarded data from
one cluster to the other or to the base station for processing.
Another issue with the ad hoc nature of WSNs is
transmission interference. The MAC protocol must mini-
mize both power usage and data broadcasting collisions.
WSNs use IEEE 802.15.4, which uses carrier sense multi-
ple access with collision avoidance multiplexing. This allows
multiple nodes to access the same channel at different
times without interference. However, the 2.4-GHz ISM band
used by IEEE 802.15.4 is crowded causing variable and
unpredictable transmission. Therefore, Liu et al. [76] in-
vestigated frequency diversity (FD) using modified sensors
to monitor an underground railway tunnel where the con-
fined environment made data transmission difficult. In FD,
the data are transmitted using several frequency channels.
Liu et al. [76] found that applying FD improved transmis-
sion for WSNs in confined environments such as tunnels.
Shafiullah et al. combined energy efficient MAC proto-
cols [114] with multifunctional nodes [112]. They developed
energy-efficient protocols for IEEE 802.15.4/ZigBee multihop
(daisy chain) networks mounted in train carriages. The nodes
acted as both sensors and routers to both take measurements
and forward data to a base station in the train locomotive. One
MAC protocol (E-BMA) is optimal for low-medium traffic and
the second (EA-TDMA) is optimal for medium-high traffic. In
E-BMA, sensor nodes use piggybacking to reserve transmission
slots rather than sending control messages to reduce transmis-
sion and, thus, reduce energy usage. E-BMA also waits for
one extra time step to determine if there are multiple packets
to transmit and sends them together. EA-TDMA divides the
transmission channel into time slots and allocates slots to sensor
nodes to avoid collisions. The nodes wake up in their slot,
transmit if they have data or switch off immediately if they have
no data, and sleep during the remaining time slots to conserve
energy.
Salaberria et al. [108] developed an alternative multitrans-
missiontechnique using GPRS. They used two phonecards be-
longing to different telephone providers on board eachtrain to
increase the network coverage. The hardware can then switch
between networks depending on network availability.
IV. MONITORING SYSTEM
The data from sensors is treated as either a time series,
where data are produced continuously or periodically, or a
sequence of readings where data is generated ad hoc, for
example, generated every time a train passes. The data can
be monitored by searching for thresholds (triggers), known
problem signatures (classification), identifying unknown events
(short-term analysis using outlier detection), or identifying drift
over a longer period of time (long-term outlier detection).
Condition monitoring can be performed continuously or
periodically. Continuous monitoring should detect a problem
straight away but it is often expensive; energy hungry, which
is a problem for WSNs where the network components need
power; and the sensor data are very noisy, which requires
careful preprocessing to ensure accurate diagnostics. Periodic
monitoring is cheaper, uses less energy, and allows time for data
cleaning and filtering but a problem will only be diagnosed at
the next processing run. This may be acceptable for some situ-
ations that change slowly such as cracks developing in bridges
but for time critical scenarios, then continuous monitoring is
necessary.
In basic condition monitoring, the system is only able to
distinguish between normal and abnormal conditions (no fault
or fault). Beyond this, systems exhibit increasing levels of mon-
itoring sophistication as outlined by the staircase of structural
health monitoring of Lopez-Higuera et al. [78] described in
Section I. A Level 5 system will comprise complex hardware,
custom algorithms, and software to allow the diagnosis and/or
the prognosis and even the solutions.
The topology of WSNs often varies over time. One very
important factor in this topological variation is the mobility of
the sensor nodes. Lorestani et al. [79] subdivided the commu-
nications network for their WSN into two: the fixed network
relates to sensor nodes in fixed locations such as bridges, tun-
nels, and special points, whereas the movable network relates
to sensor nodes attached to locomotives or rail wagons. The
data for the movable network are logged with accompanying
GPS coordinates. Movable (on-board) sensors can monitor the
whole track length travelled by the train but only monitor the

sections of the train where the sensor nodes are attached [84].
In contrast, track-mounted (fixed) sensor nodes can measure
the whole train as it passes but only at specific points where
the nodes are mounted on the track. This tradeoff needs to be
considered when designing the node placements. The following
sections analyze existing WSNs for condition monitoring on
the railways and conducts analyzes from the perspective of
fixed (static) and movable (dynamic) monitoring [122]. In each
category, the systems are grouped according to the object(s)
being monitored.
A. Fixed Monitoring
Many authors have investigated health monitoring of infras-
tructure, including rail bridges, tunnels, rail tracks and track
beds, and other track infrastructure. Krüger et al. [70] noted
that structural health monitoring is becoming ever more impor-
tant as both train speeds and the axle loads exerted by trains
increase. Until recently, inspection has been performed visually
[70], but this only examined the structure superficially and
intermittently. Additionally, the visual analysis needs to be in-
terpreted by an expert, which can be subjective. Sensors are ob-
jective and can provide data from all of the structure (including
internally) to allow the whole structure’s health to be assessed
and to analyze its durability and remaining life time. Wired
sensors can be used for monitoring. However, wired systems are
expensive, inflexible, time consuming to install, and the trains
have to be stopped during installation [70]. The early fixed
WSNs were simple such as attaching sensor nodes to the rails
to monitor the rail temperature or low-voltage warning sensors
that monitor the power supply to motors at points [15]. These
systems simply generated a binary output (high or low). For the
temperature sensor, a rail expert with knowledge of the ambient
(stress free) rail temperatures set the alarm levels for “amber”
and “red” warnings. If an alarm level was reached, then the
system sent an SMS alert. Modern WSNs provide semiau-
tomatic or automatic analysis of the sensor data to examine
structural changes and to improve the durability of structures.
Hence, WSN monitoring should reduce the overall maintenance
costs. An overview of fixed monitoring systems is given in
Table II in the Appendix.
1) Bridges: Bridges suffer structural defects exacerbated by
the constant strains and vibrations of passing trains. Human
inspection of bridges is difficult and much of the structure may
be inaccessible. WSNs enable constant monitoring of the whole
structure, including the internal structure of the bridge. The
monitoring systems often comprise two units, as detailed in
Section II-C. The first unit is low power. It detects the approach
of a train and wakes the second unit. This second unit generates
the measurements, whereas the bridge is under load (a train is
crossing) to assess the bridge’s health.
For an overall assessment, Krüger et al. [70] used a variety
of MEMS to monitor structural changes in concrete including:
piezo-based acoustic emission (AE) sensors for crack and
fatigue detection, strain gauges to analyze static loads and
stresses, accelerometers to analyze dynamic loads as trains
pass, and temperature/humidity sensors to determine the am-
bient conditions. Grosse et al. [53], [54], Hay et al. [58], Hay
[59], Ledeczi et al. [73], and Nair and Cai [90] used AE testing
to detect fatigue cracks and other internal structural damage in
bridges. The rate and intensity of AE from cracks defines the
crack growth rate, which Hay et al. [58] categorized as inactive,
active, or critically active. Ledeczi et al. [73] combined AE
sensors to detect crack growth with strain gauges to identify
changes in the forces exerted by trains both of which indicate
structural problems. Bischoff et al. [19] and Feltrin [46] also
analyzed force changes using resistance strain gauges to detect
structural damage. Similarly, Sala et al. [106] and Sekula and
Kolakowski [109] analyzed force changes using piezoelectric
strain sensors mounted on the bridge trusses. Kolakowski et al.
[67] used ultrasonic strain gauges to record elastic waves
in bridge trusses due to vibrations caused by passing trains.
Changes in vibrations can indicate structural problems simi-
lar to changes in forces. Chebrolu et al. [32] detected these
changes in vibration using accelerometers. During monitoring,
Kolakowski et al. [67], Sala et al. [107] and Sekula and
Kolakowski [111] used an extra module: piezoelectric strain
(“weigh in motion”) sensors mounted on the track approaching
the bridge, which weigh the train; thus, the train’s weight can
be factored into the structural assessment.
Kerrouche et al. [65] incorporated optical sensors into the
fabric of bridges during bridge repairs to allow structural anal-
ysis. They used fiber Bragg grating (FBG) strain sensors with
grooves cut in the bridge fabric to house the optical fibers
carrying the FBG sensors. The authors concluded that the FBG
monitoring needs more thorough evaluation as it was difficult
to distinguish the actual strain from the noise.
2) Tunnels: Tunnels can be damp leading to rapid corrosion
damage and other defects; thus, monitoring is essential. Mon-
itoring of tunnels has many similarities to bridge monitoring.
However, rail tunnels are often difficult to access and inspect.
An added complication of tunnel monitoring compared with
bridge monitoring is data transmission in a confined environ-
ment [88]. Data generated by sensors inside the tunnel have to
be transmitted using a suitable relay mechanism (often multiple
hop using relay nodes) from the various sensor nodes to the out-
side. Bennett et al. [17], Cheekiralla [33], Hoult et al. [62], and
Stajano et al. [116] described WSNs deployed in the London
Underground tunnels. Bennett et al. [17], Hoult et al. [62], and
Stajano et al. [116] detected distortion using inclinometers and
measured movement using linear potentiometric displacement
transducers. They coupled the sensors with four relay nodes
in a mesh topology to transmit the sensor data to a receiver
unit in the tunnel. There is no GPRS signal near the London
Underground tunnels. The authors used repeater units placed
along the tunnel to transmit the data from the gateway to a
computer located outside the tunnel at the top of a ventilation
shaft. This transmission computer logged the data and trans-
mitted them to the offsite compute server using a mobile-phone
signal. Bennett et al. also analyzed tunnels in the Prague metro
using the same sensor network. Here, a GPRS signal is available
at the tunnel mouth; thus, they connected the gateway node to
the transmission computer directly using an Ethernet cable with
no requirement for repeater units, thus, reducing the chances
of data loss. Cheekiralla [33] monitored deformation in the
tunnel when construction work was happening nearby using
off-the-shelf battery-powered pressure transducers to measure

vertical displacements and the transverse deformation along
the most vulnerable zone of the tunnel. Cheekiralla coupled
each pressure transducer with a microcontroller, RF transmitter,
analog-to-digital converter, and power conditioning hardware.
The base station was located near the tunnel with a RF receiver
unit connected to a laptop. The laptop had a wireless modem to
periodically send data to a server.
3) Rail Tracks: Track monitoring systems also play a vital
role in maintaining the safety of the railways. Monitoring
bridges and tunnels (discussed in the previous paragraphs) uses
sensors to identify and analyze defects (cracking and displace-
ment) in large structures. In contrast, track monitoring involves
identifying and analyzing defects in long narrow metal rails.
Tracks can crack and displace like bridges but also twist and
tilt (incline). Hence, track monitoring can vary from detecting
settlement and twist such as that caused by nearby tunneling
or excavation [112]; to measuring the forces exerted by train
wheels on the tracks; to monitoring the development of cracks
and structural flaws as trains pass and over the longer term.
Durham Geo Slope Indicator Company (DGSIC) [112] and
Aboelela et al. [1] analyzed the track movements (shift) over
time particularly where nearby construction work may affect
the track or where track passes through vulnerable areas such
as regions prone to landslides. Aboelela et al. [1] mounted
inclinometers parallel with the tracks directly on the sleepers
or in the ballast with continuous tensioned rails to monitor tilt.
Likewise, DGSIC [112] mounted inclinometers parallel with
the tracks to measure settlement. They also mounted inclinome-
ters perpendicular to the track (on the ties) to monitor twist.
Aboelela et al. [1] also plan to use inclinometers to monitor ties
to detect broken ties or sabotage.
Other systems analyzed the effects of trains passing along
the tracks using track-mounted strain gauges [10], [71], [94],
[99], piezoelectric strain sensors [112], FBG strain sensors
[41], [72], [118], [119], [127],AEsensors[7],[22],[121],[128],
[135], and accelerometers [10], [18], [71]. Current sensors
readings can be compared with historical sensor measure-
ments and unusual readings or readings that match known
problem signatures can be detected [7], [22], [41], [72], [118],
[121], [127]. Strain gauges monitor the forces exerted by trains
(“weigh in motion”), and accelerometers measure vibrations
(motion of the track [9]), whereas AE sensors detect and
analyze cracks. Tam et al. [119] installed FBG strain sensors
on each side of a track to detect imbalances on the two sides of
train wheels. If there is a large variation between the left and
right axle loading then there is the danger of train derailment.
Monitoring the impact of passing trains is important for
two reasons: it can detect track damage, but it can also de-
tect damaged trains. Brickle et al. [23] detailed state-of-the-
art automated systems using track-mounted sensors for wheel
condition monitoring. Excessive forces, particularly where the
train has a defect such as out-of-round wheels, flat wheels, or
unbalanced axle loading, can crack sleepers, damage the rail
head, and cause failure of rail by either growth of fractures, or
fatigue and fracture of in-track welds [10]. Filograno et al. [47]
used FBG sensors for train identification, axle counting, speed
and acceleration detection, wheel imperfection monitoring, and
dynamic load calculation by placing them in different positions
on the track. Monitoring the forces exerted using strain sensors
can identify defective axles and wheels [118] or incorrectly
loaded trains [41]. Analyzing the track vibration data produced
by accelerometers [16] and the AE signal transferred through
the rail [7], [22], [121] from AE sensors can also detect out
of round wheels as trains pass. Belotti et al. [16] incorporated
an inductive axle-counter block into their system for assessing
the train speed. The speed affects the expected vibrations; thus,
the speed can be built into the wheel analysis model. Detecting
wheel faults can also pinpoint trains that are at risk of causing a
derailment. In [10], [71], the risk of derailment is represented by
the ratio between the lateral and the vertical force (L/V ratio),
which quantifies the relationship between the applied load and
the deflection of the track [10].
Condition monitoring protects both the trains and the track,
increases the track and train reliability and allows repair to
be scheduled. If the track has been marked as defective and
awaiting repair but is still usable, AE sensors can be attached to
monitor the defects and ensure that they are not getting worse
[135]. Track awaiting repair usually has a speed restriction
enforced so, if it is monitored, this may not be required.
4) Rail Track Beds: As well as monitoring the track, it is
important to analyze the rail bed [8]. The pore-pressure of water
in the subsoil of rail beds increases as trains pass. Repeated
pressuring can reduce the shear strength of the soil and increase
the potential for bed failure. Wong et al. [132] used vibrating
wire piezometers to measure this pore pressure. Vibrating wire
piezometers are accurate and reliable. However, they have a
limited reading frequency due to the vibrating wire. Wong et al.
[132] generated data at a rate of 0.1 Hz and Hendry [60] notes
that this low rate of data acquisition may miss the peak pore
pressures under axle loads which occur at frequencies of >1 Hz.
Konrad et al. [68] measured pore pressure using piezometers
and wire potentiometers which generate data at >100 Hz.
This higher frequency allows the analysis of pore pressure
during the passage of individual trains. Other systems use a
variety of sensors for monitoring. Aw [8] analyzed a railway
site experiencing mud pumping problems that requires frequent
(bimonthly) track maintenance. Aw used accelerometers to
monitor the track vibrations, settlement probes to monitor track
settlement, piezometers to measure ground water pressure, and
temperature sensors. Hendry [60] used piezometers to measure
the pore pressure and detect changes; extensometers to measure
the relative vertical motion at various depths; and, an arm
comprising jointed segments with accelerometers attached. The
movement of the segments and overall shape of this arm is used
to analyze the horizontal deformation of the foundation material
during the passage of a train. Baum et al. [14] developed a WSN
early warning system for rail landslides. It combined data from
rain gauges to determine the rainfall level and likely saturation,
tensiometers to measure the pore pressure and reflectometers to
measure the soil water content. Tensiometers are able to detect
changes in pore pressure rapidly to warn of slope instability,
whereas reflectometers detect water content changes over time
to warn of wet soil slippages.
5) Other Track Infrastructure: A variety of infrastructure
is used to operate the tracks. Monitoring the condition of
this infrastructure is an important part of railway condition

monitoring. Fishplates are metal plates attached to the end of
track sections to join and align them [51]. Fishplates can be
tampered with, incorrectly maintained or the nuts and bolts
may work loose. This may cause track sections to become
misaligned potentially resulting in a derailment. Buggy et al.
[27] and Ghosh et al. [51] proposed WSNs to monitor fish-
plates using FBG stain sensors [27] and piezo-resistive pressure
sensors [51], respectively. If the nuts and bolts of a fishplate
are loosened, then the strain and pressure on them change. The
WSN can send an alert to the nearest control room.
The railways use track circuits which detect approaching
trains and ensure correct train and signal operation [36], [49],
[133]. Track circuits transmit a signal to a receiver via the two
running rails of a track section [36]. If the signal reaches a
receiver at the far end of the track section, then no trains are
present. If a train is present on the section, then the signal is
short circuited by the wheels, the signal does not reach the
receiver and the track relay is not energized. Fukuta [49] noted
that this receiver can be a wireless MEMS. Detecting current
flows along the track can also pinpoint broken rails or discon-
nected wires. Railway points switch (move) the track between
two alternative routes to ensure the train takes the correct route
[63], [83], [86], [136]. The points work in conjunction with
track circuits to detect and route trains. Hence, both the track
circuits and the points must be fully functioning to accurately
detect and route trains to ensure safe and reliable operation.
Points monitoring systems often use a variety of sensors as
they require multifaceted monitoring. Zhou et al. [136] mon-
itored: distance, motor driving force, driving current and volt-
age, electrical noise, temperature and state changes, whereas
Marquez et al. [83] measured a similar set of variables: voltage,
current drawn by the electric motor, current drawn by the
system as a whole, and the force in the drive bar measured with
a load pin. Igarashi and Siomi [63] used magnetic lock-warp
sensors to detect maladjusted switches and points irregularities
by measuring displacement through temperature variations.
B. Movable Monitoring
WSNs are able to monitor the condition of a range of
mechanics, systems, and environments using on-board sensors
to measure parameters such as temperature, shocks, tilts, and
humidity [138]. Thus, WSNs demonstrate a promising platform
for real-time movable condition monitoring of transport sys-
tems, allowing the early detection and diagnosis of problems.
Bruni et al. [24], Ngigi et al. [95] and Ward et al. [126]
provided overviews of the opportunities available for on-board
vehicle monitoring systems, which analyzed the mechanics.
Measurements can all be taken during normal train service that
allows in-service condition monitoring of the train chassis and
mechanics and the rail track that the train is running on. The
sensor nodes used for movable monitoring are often mounted
on the train bogies or carriages. These locations are susceptible
to high levels of vibration; thus, the sensor nodes have to be
rugged or protected. Authors use strong adhesives to bond the
nodes to the object being monitored and sealed protective cases
to shield the sensor nodes. For example, Grudén et al. [55] used
sensors to monitor train bogies. They used shock absorbing
material around circuit boards to protect them, separated the
battery from the circuit board, mounted the antenna separately
from the circuit board, and used flexible cables.
One of the key issues with movable sensor monitoring is
communication, due to the mobility of the sensors. Movable rail
monitoring can transmit the data from the movable nodes to a
static node when the static node is in range. Transferring data
from the movable nodes to a static node often happens when
a train stops in a station. Alternatively, the movable nodes can
form a network with a base station within the network. If the
nodes form a network on the train, then the base station is often
mounted on the locomotive, where power is available to power
the data transmission. This transmits the data over a suitable
mechanism, such as satellite or GSM. Despite the issues with
movable monitoring, mobile sensor nodes can achieve a higher
degree of coverage and connectivity compared with static sen-
sor nodes. As they move to a range of locations.
Object mobility (such as the movement of sensors) can be
classified into uncontrolled and controlled mobility [13]. With
uncontrolled mobility, a sensor node can move randomly and
freely. In contrast, a node only moves along a predetermined
route with controlled mobility. Vehicles on the railways can
only move in limited trajectories, each train follows a planned
route along tracks, which are set by points and signals and
the movements are dictated by the rail timetable. Establishing
WSNs for controlled mobility is easier compared with WSNs
for uncontrolled mobility. If the routes are known and fixed,
then the nodes and receivers can be arranged for maximum
efficiency as wireless sensor nodes have limited transmission
ranges. However, Chen et al. [34] noted that even with con-
trolled mobility, trains can be over a mile long. If these trains
are fitted with sensor nodes, then the off-train receivers have
to be carefully designed and positioned to cope. An overview
of movable monitoring systems is given in Table III in the
Appendix.
1) Train Engines, Passenger Carriages, and Freight Wag-
ons: Wagon or container failure could lead to damage of the
rail infrastructure or the environment and could even result in
loss of life; thus, it is important that the wagons, containers,
and their respective contents are also monitored. Additionally,
freight trains use unpowered and unwired railroad wagons, thus
an on-board WSN can analyze the condition of the freight wag-
ons [104] without power from the train. A number of authors
have developed systems for analyzing train carriages, wagons,
and engines. This can involve analyzing the mechanics of the
vehicles or analyzing the contents where safety-critical material
is being transported. Many of these systems form a mobile
network in a daisy chain topology between adjacent carriages
and with the engine acting as the base station [50], [81], [104],
[112], [131]. The sensor data can also classify the rail wagon’s
response to the track and even detect potential derailments.
Ward et al. [126] noted that on-board sensors can identify track
defects including small vertical irregularities (dipped joints
or voids), small lateral irregularities (kinks), larger vertical
irregularities, larger lateral irregularities (misalignment), and
track twist.
A number of authors analyzed the mechanics of carriages
and engines using parameters such as the temperatures, strains,

and forces at critical locations on the rail wagon shell and
mechanics. Gao et al. [50], Nejikovsky, and Keller [93]
and Tam et al. [119] used WSNs to monitor the dynamics
and structural condition of both locomotives and passenger
carriages, whereas Shafiullah et al. [112] and Wolfs et al. [131]
monitored carriages. In [93], engine sensors analyzed the health
of locomotive engines by measuring parameters, including
water temperature, oil temperature, engine temperature, switch
statuses, fans statuses, throttle positions, rotations per minute,
and traction motor currents. Similarly, Tam et al. [117] used
FBGs to analyze the strain and temperature of train body shells
in critical locations, which include the welding joints, cross
beams, and sole bars. Accelerometers can analyze different
properties of the wagons depending where they are mounted
[50], [93], [131]. Wolfs et al. [131] placed them at the edges of
rail carriages; Gao et al. [50] placed them on the floor of the
locomotive and the chassis of carriages; and Nejikovsky and
Keller [93] mounted them on the body of passenger carriages of
tilting trains to monitor rail wagon body motions. These train-
mounted accelerometers can measure vertical and horizontal
vibrations [50], [131]. They can also monitor the lateral acceler-
ations when travelling around curves to ensure the lateral forces
do not exceed government safety levels [93]. Gyroscopes [131]
can measure the longitudinal accelerations, as well as pitch,
roll, and yaw of carriages to analyze motion.
Casola et al. [31], Flammini et al. [48] and Maly et al.
[82] proposed to integrate data from a range of heterogeneous
sensors for more holistic condition monitoring. Maly et al.
[82] performed hot box detection, flat wheel detection, dynamic
scaling, and loading gauge detection using commercially avail-
able sensors, whereas Casola et al. [31] analyzed temperature,
humidity, and vibrations among others (the specific sensor
types are not stated). The sensors are grouped into units, which
operate independently. If a unit detects a faulty train, then it
alerts the control center and provides detailed information about
the detected irregularity.
Wagons containing freight are susceptible to tampering and
intrusion. Authors have combined accelerometer data with
light, shock, humidity, and temperature readings to provide
security monitoring [81], [104], [112]. Mahlknecht and Masani
[81] used vibration shock sensors; Mahlknecht and Masani
[81] and Shafiullah et al. [112] monitored the humidity to
detect changes; Mahlknecht and Masani [81] and Reason et al.
[104] used light sensors to detect wagons that are open, and
Mahlknecht and Masani [81] used motion detectors for analyz-
ing movement and detecting irregularities.
2) Train Bogies: The train bogies perform a number of
tasks, mainly to guide the train on both straight and curved
tracks, to safeguard the train’s dynamic stability, and to provide
a comfortable ride for the passengers [126]. Most commercial
condition monitoring systems for rail vehicles analyze the bogie
system. Ngigi et al. [95] noted that this is because the bogie’s
important components undergo rapid changes during journeys
and can pose safety related issues. Train axles and bogies run
hot during train journeys. However, excess heat is generated
as the axles and bogies wear. It is imperative to detect this
excess heat to allow early detection of wear and deterioration,
schedule maintenance and thus prevent vehicle failures and
derailments. Hot axle bearing detectors (HABDs) have been
in use for a number of years [98]. The early HABDs could
detect defects but were only capable of generating an alarm
once a certain threshold was reached [98]. In contrast, new
systems generate a significant amount of data, both as alarms
and as time-series graphs of the bearing temperatures. Different
heat types have different time-series signatures [98]. Hence,
it is possible to identify faults such as hot wheels caused by
brake drag or extraneous heat sources such as leaking engine
exhaust fumes. Previously, when only a simple threshold was
used, alarms caused by these other sources of heat may have
been misinterpreted as false positives, due to the lack of detailed
information [98]. Rabatel et al. [102], [103] and Carrascal et al.
[30] used MEMS temperature sensors; Kim et al. [66] used
surface acoustic wave (SAW) sensors mounted on the train’s
bogies to detect overheated wheel bearings and Grudén et al.
[55] used temperature sensors to analyze wheel performance.
Grudén et al. [55] used three sensors to measure bogie tem-
peratures and a fourth sensor to measure the air temperature
to establish the ambient conditions. The ambient conditions
affect the train mechanics, thus the need to be incorporated into
the condition monitoring to prevent false positives and false
negatives.
Bogies can also be analyzed for vibrations. Elia et al. [44],
Mei and Li [87] and Tsunashima et al. [124] proposed sys-
tems for the measurement of bogie vibrations. Elia et al. and
Tsunashima et al. mounted accelerometers on the bogies and
Elia et al. also placed them on axle boxes of train carriages to
measure the lateral acceleration. Mei and Li used robust inertia
sensors mounted onto the bogie frame to monitor vibrations
and ground speed. The papers also demonstrated that rail/wheel
conicity caused by incorrect track gauge or inclination [44]
and track irregularities [87], [124] can be identified from the
respective bogie-based sensors. Tsunashima et al. [124] were
able to analyze the sensor data onboard the train at source to
provide real-time track analysis.
3) Train Axles: As well as generating excess heat, excess
strains (such as torsional and bending stress) may be exerted
on axles causing deterioration. The degree of torsional stress
is affected by the train’s traction control system, mechanical
brake characteristics, the wheel–rail interface and the track
adhesion conditions. This number of factors affecting the stress
means that establishing a baseline to create a model necessitates
measuring the torsional axle stress levels over a long term,
ideally over the four seasons of a year [42]. Similarly, the
axle bending stress levels vary due to the track quality, vehicle
loading, braking and traction effects and curving speeds. Deuce
and Rosinski [42] also observed strong electromagnetic inter-
ference (EMI) when measuring strains on powered axles. This
required the development of bespoke instruments, including
strain gauge amplifiers, telemetry systems, solid state data
loggers, and signal processing units equipped with wireless data
communication [42].
4) Train Wheels: A number of the systems already dis-
cussed for track and train vehicle monitoring are able to detect
wheel irregularities. Train wheels wear over time and can
also suffer flat wheels and out-of-round wheels. The previous
section (Section IV-A fixed monitoring) discussed how WSNs

attached to the track detected faulty wheels by analyzing vibra-
tions and stresses exerted by the wheels on the tracks. Wheel
problems can also be detected by sensors on board the train
analyzing the wheel surfaces or analyzing forces on the train
bogies and wheels. The wheel-sets and bogie each have two
degrees of freedom (lateral and yaw) [75]. Wired sensors could
be used for analysis but require holes in the train axle for the
power leads, which weaken the axle, whereas wireless sensors
are not very durable in the harsh conditions [84]; thus, this lack
of durability has to be considered.
Nuffer and Bein [96] introduced a wear detection system for
train wheels. It detected changes in the vibration behavior of the
entire wheel caused by surface changes on the rolling contact
area. The system positioned piezoelectric sensors on distinct
areas of the wheel. The sensors represented the surface as an
electrical signal. Changes in the surface can then be detected
by changes in the electrical signal. However, it proved difficult
to adequately define the correlation between the behavior of the
wheel and the measured signals; thus, the system was sensitive
to small changes in model parameters. Li et al. [75] noted
that the behavior models for rail monitoring are very complex
with strong interdependencies between the measured variables
and between different degrees of freedom. Thus, a number of
variables need to be measured. They mounted sensor nodes
on the bogie frame to monitor the dynamic behavior of wheel
profile (conicity) and dampers. Two accelerometers mounted
on the front and rear wheel set measure the lateral accelerations
of the wheel set. An accelerometer and a gyromounted on the
bogie measured the lateral acceleration and yaw velocity of
the bogie. Finally, a vehicle body accelerometer measured the
lateral acceleration of the vehicle body.
It is also important to monitor the wheel/rail contact to
detect skidding and wear. Matsumoto et al. [84] applied a
multifaceted approach. They fitted noncontact gap sensors on
nonrotating parts of a bogie to measure the bending deflection
(distortion) of the wheel by the lateral contact force. They
also used magnetostrictive displacement sensors to measure the
deflection of the primary suspension by the vertical contact
force. These two measurement combined give the wheel/rail
contact forces. Matsumoto et al. [84] used the same derailment
coefficient as [10] and [71] (discussed in the fixed monitoring
section): the ratio between the lateral and the vertical force (L/V
ratio). Balas and Jain [9] fused data from speed sensors and
piezoelectric accelerometers attached to train wheels. Using
data modeling, they were able to detect skidding if the integral
of the acceleration (speed of the wheel) measured by the
accelerometer differed from the speed of the carriage. They also
calculated the wheel adherence as a function of the carriage
weight, carriage velocity, brake cylinder pressure, and the skid
amount from the skid model.
5) Train Brakes: Heavy haul transportation involves car-
rying heavyweight railway freight. The trains have multiple
locomotives, which must coordinate their braking. Song et al.
[115] developed a WSN for brake coordination and monitoring
to reduce the maintenance and inspection requirements. They
attached air pressure sensors to the locomotives and train rear,
installed relay nodes on the carriage roofs, and placed ac-
celerometers and ultrasonic sensors along the track along with
railway relay nodes. When the front locomotive brakes, the sen-
sors in it detect the change of air pressure, and the relay node on
the locomotive transmits a message to the other locomotives via
the relay nodes on either the train or the track using a wireless
radio link. The other locomotives can then apply their brakes
according to the message from the lead locomotive. The track
relay nodes are used where transmission via train carriages is
weak such as bends and tunnels. The data regarding the railway
infrastructure from the accelerometers and ultrasonic sensors
along the track can be collected and analyzed for condition
monitoring.
6) Train Pantographs: In electrified railways, the overhead
power lines and the pantographs on the trains require monitor-
ing. The pantograph ensures that the train contacts the overhead
power line to collect current and power the train. Ensuring
the integrity of the pantograph, the pantograph/overhead line
interface and the overhead power lines is important to ensure
the power supply to the train, prevent breakdowns and identify
sections of power line requiring maintenance. Thus, monitoring
the force exerted between the contact and the wire is essential
to detect and prevent problems.
Boffi et al. [21], Bruno et al. [25], Elia et al. [44], Waki [125],
and Willsch et al. [129] all performed condition monitoring of
the pantograph itself. Elia et al. also monitored the overhead
power lines for irregularities and loss of tension. Optical sensors
are immune to EMI making them ideal to use near power lines
where EMI is high. Thus, Boffi et al. and Willsch et al. mea-
sured the strain exerted between the pantograph’s current col-
lector and the overhead wire using interferometric fiber [21] and
FBG [129] sensors. Elia et al. and Waki used a variety of sen-
sors to analyze different aspects of the pantograph. Both used
accelerometers that can detect impact accelerations that occur
when passing a contact failure and both integrated sensors in the
current collector. Elia et al. used load cells to measure the reac-
tion forces and displacement transducers to measure the motion
of the current collector and articulated frame. Waki along with
Bruno et al. attached ultraviolet (UV) sensors to the current
collector to detect arcing. Arcing occurs when the contact
between the pantograph and the overhead wire is lost and causes
power transmission loss and overheating. To detect overheating
further, Waki measured the pantograph’s temperature using a
standard temperature sensor. Waki used a video camera to
record overhead line conditions and pantograph movement.
Leung et al. [74] used passive magnetoelectric current sensors
to analyze the current in the transmission cable running from
the pantograph to the train’s power unit. Changes in the current
flow can indicate power transmission loss and need to be
detected.
V. CONCLUSION
This paper has reviewed the range of WSNs used for con-
dition monitoring in the railway industry. The emphasis is on
the practical engineering solutions, principally which sensors
devices are used and what they are used for; and, identification
of sensor node configurations and network topologies.
Until recently, railway inspection has been visually per-
formed [70], but this only examines objects superficially and

intermittently, and the analysis needs to be interpreted by an
expert, who can be subjective. Sensors are objective and can
provide data from the entire object (including internally) to
allow the whole object’s health to be fully assessed and to
analyze its durability and remaining life time. A broad range of
sensors are used in railway monitoring to provide an extensive
range of data and allow monitoring of different structures,
vehicles and machinery. The main challenge for WSNs in
railway applications is determining the best measurement tech-
nologies to use. The WSN must be reliable and accurate to
enable effective condition monitoring in harsh and inaccessible
environments but must also be cost effective. It must be possible
to translate the sensor data from the WSN into relevant and
clear information to enable decision support in the railway
infrastructure maintenance lifecycle.
The paper divides railway condition monitoring into fixed
monitoring for immobile infrastructures such as bridges, tun-
nels, tracks and associated equipment, and movable monitoring
for vehicles and their mechanics. Fixed monitoring uses sen-
sors to monitor vibrations, stresses and sound waves passed
through structures (acoustics) caused by passing trains (short-
term monitoring) and also changes in stresses, pressures and
sound waves passed through structures over the longer term
(long-term monitoring).
One of the key issues for fixed monitoring is network topol-
ogy. The topology is constrained by the requirements of the
monitoring and by the physical environment. Sensor nodes can
be arranged in either an ad hoc or a preplanned configuration
[134]. Determining the optimum node placement is a complex
task and often requires a tradeoff. The network configuration
can be optimized against a number of different constraints. A
network may minimize relay nodes, may need to ensure a min-
imum level of service (include a certain level of redundancy),
minimize energy usage to preserve battery life, or may need
to ensure accessibility of the nodes [116]. The communication
mechanisms, for example, Bluetooth, Wi-Fi, GSM, or satellite
also need to be evaluated to ensure coverage and reliability and
the routing protocol to ensure data is successfully transmitted
from sensor node to base station to control center. Another
issue is powering the sensors as fixed monitoring often requires
placing sensor nodes in inaccessible locations, for example, in
tunnels, on bridge trusses, or in rail track beds. Many sensors
use batteries but replacing batteries in inaccessible locations
may not be possible. Authors have considered ambient energy
harvesting such as converting vibrations caused by passing
trains into energy or using solar power. These energy harvesters
may be accompanied with energy storage such as capacitors.
Authors have also reduced sensor energy usage using event
detection where a single cheap sensor detects approaching
trains and wakes a larger sensor network to commence taking
measurements of a bridge structure.
Movable monitoring uses sensors to monitor temperatures,
vibrations, stresses, forces, currents, and sound waves in vehi-
cles and their mechanics particularly bogies and chassis. These
locations are susceptible to high levels of vibration and EMI;
thus, the sensors have to be rugged or protected. One of the key
issues with movable sensor monitoring is communication due
to the mobility of the sensors. Movable rail monitoring either
transmits the data from the movable nodes to a static node when
the static node is in range, or the movable nodes form a network
with a base station within the network which transmits the data
over a suitable mechanism, such as satellite or GSM.
This paper focuses on the sensor technology used to generate
condition monitoring data to enable practical condition moni-
toring systems. These data must be managed and turned into
useful information to generate useful information. Condition
monitoring systems must store large quantities of data to build
models for analysis. The data must be validated first to ensure
that they are correct and error-free (sensor faults, noise, null
values, communication errors, etc.). This process may even be
performed in the sensor node’s microcontroller; thus, only valid
data are transmitted thus minimizing the transmission load.
These data can then be processed in a number of different ways
to generate information. Once data are collated, they can be
analyzed using robust algorithms to identify faults in near real-
time. Algorithms need to be robust as WSN data is noisy, can be
intermittent, may contain errors, has many interdependencies
and the data volume is very high. The WSN data can also be
stored and analyzed over longer time periods to identify long-
term progressive faults such as a slowly developing crack.
Some systems incorporate contextual data that describe the
ambient conditions, which will affect the object monitored.
These contextual data can be built into models to improve
the coverage and accuracy of the model and to help provide
explanation of condition monitoring decisions. Elia et al. [44]
included: train speed, position along the line (longitude and
latitude from GPS system) and wind speed and direction when
analyzing bogie vibrations, whereas Rabatel et al. [102], [103]
included structural criteria (route) and environmental criteria
(weather, travel characteristics, travel duration) when analyzing
bogie temperatures.
VI. FUTURE RESEARCH
There are a number of promising directions for further
research in condition monitoring in the railways. One future
direction is a move toward holistic integrated systems [31],
[48], [82] which provide near real-time information and alerts.
These approaches will integrate data from different sensor sys-
tems using sophisticated modeling techniques. They will also
incorporate more contextual data into the modeling, including
the ambient conditions, the route, the journey time, the weight
of the train, and more. The models will use multistage data
fusion. This fuses measurements from a variety of sources
(sensors and contextual data) and over a range of time epochs
to generate a consolidated state history of the object being
monitored [38]. The models will also mitigate data depen-
dencies (physical dependencies) across the physical objects
being monitored. An integrated multifaceted approach should
improve prediction quality. For example, train monitoring data
can be combined with route data and GPS data. If multiple
trains detect a vibration fault at exactly the same GPS location,
then the fault is more than likely in the rail infrastructure such as
uneven track. However, if only one train identifies a vibration
anomaly at a particular GPS location, then the fault is more
likely to be in the rolling stock [64].

TABLE I
TABLE DETAILING THE SENSOR DEVICES USED IN RAILWAY CONDITION MONITORING WITH WSNS
As sensors become cheaper and easier to fabricate, then
their availability and use will increase. Bukova and Svecova
[28] proposed a WSN comprising many tiny wireless MEMS
(smart dust), which are able to collect, process, and transmit
data from light to vibrations. Vast numbers of these sensors
can be placed inside a wagon, for example, where they can

TABLE II
TABLE LISTING THE FIXED OBJECTS MONITORED, THE MEASUREMENTS USED FOR EACH OBJECT, THE SENSOR DEVICES
USED TO MONITOR IT AND A LIST OF CITATIONS TO SYSTEMS MONITORING THAT OBJECT
measure temperature, humidity, or vibrations. Bukova and
Svecova posited that recent developments in sensor fabrication
techniques will produce MEMS that are the size of a grain
of sand yet contain computing circuits, bidirectional wireless
communications technology and a power supply.
The number of sensors will increase in the future, and the
number and distribution of monitored objects will expand. This
will require further development of energy efﬁcient routing
and transmission mechanisms to ensure reliable and consistent
communication. Condition monitoring will also tend to parallel
and distributed processing across multiple nodes to speed exe-
cution. This could be processing the data in parallel at the same
geographical location using a single machine with multiple
processing cores or at the same geographical location using
multiple compute nodes or even distributed processing, where
the data are processed at multiple geographical locations, and
the results are assimilated at a central location.
As WSNs expand and become more pervasive, security
against intrusion, eavesdropping, data tampering and unautho-
rized control will be ever more important [116]. WSNs pose
unique problems for security. WSNs broadcast data between
nodes making them vulnerable. The sensors are energy con-
strained with limited processing capability; thus, any security
software needs to be carefully designed. Network administra-
tors will have to identify which WSNs need to be secured, how
they need to be secure, and what level of security is required
[117]. The type of security is important. Does the WSN need
to be secured for access? How often and to what level should
the WSN be monitored for intrusion detection? Does the data
need protecting from data tampering? The level of security will
be dictated by the value of the asset, the type of asset and the
consequences of a failure. WSNs will become more interlinked;
thus, it is imperative that WSN security cannot be compromised
through access from other networks.
In the future, ever more data will be collected from railway
infrastructure and vehicles. This inevitably leads to a require-
ment for the railway industry to develop standards for data
collection [126] and processing. West et al. [110] proposed
developing Web Ontology Language standardized ontologies to
encapsulate railway data and ensure consistency for transmis-
sion and processing. Different systems and different companies
can then share data using a consistent interface.
As more and varied condition monitoring systems are de-
veloped, then there is a further requirement of standardiza-
tion of presentation of decisions and information across these
systems for consistency and to allow integration of multi-
ple systems. Equally, many condition monitoring algorithms
are “black boxes” providing little or no explanation of deci-
sions [110]. In the staircase of structural health monitoring of
Lopez-Higuera et al. [78], a Level 4 system can estimate

TABLE III
TABLE LISTING THE MOVABLE OBJECTS MONITORED, THE MEASUREMENTS USED, THE SENSOR DEVICES USED
AND A LIST OF CITATIONS TO SYSTEMS MONITORING THAT OBJECT
the consequences of the damage and remaining service life.
However, only a Level 5 system will allow the diagnosis and/or
the prognosis and recommendation of a solution to the problem.
Future systems will need to provide clearer and more detailed
explanation of their recommendations to increase usability.
In the future, condition monitoring will be able to exploit
cheaper, more robust and more pervasive hardware. New data
processing techniques will generate more accurate, robust,
and reliable models from existing and new sensor data. Data
and processing will be standardized. Systems will be secure and
provide clear and detailed decisions and recommendations.
APPENDIX
See Tables I–III.
ACKNOWLEDGMENT
The authors would like to thank the reviewers for their
helpful suggestions, which improved the quality of this review.
REFERENCES
[1] E. Aboelela, W. Edberg, C. Papakonstantinou, and V. Vokkarane, “Wire-
less sensor network based model for secure railway operations,” in Proc.
25th IEEE Int. Perform., Comput. Commun. Conf., Phoenix, AZ, USA,
2006, pp. 1–6.
[2] M. Aguado et al., “WiMax on rails: A broadband communication ar-
chitecture for CBTC systems,” IEEE Veh. Technol. Mag., vol. 3, no. 3,
pp. 47–56, Sep. 2008.
[3] B. Ai et al., “Challenges toward wireless communications for high-speed
railway,” IEEE Trans Intell. Transp. Syst., vol. 15, no. 5, pp. 2143–2158,
Oct. 2014.
[4] K. Akkaya and M. Younis, “A survey on routing protocols for wire-
less sensor networks,” Ad Hoc Netw., vol. 3, no. 3, pp. 325–349,
Dec. 2005.
[5] I. Akyildiz, W. Su, Y. Sankarasubramaniam, and E. Cayirci, “Wireless
sensor networks: A survey,” Comput. Netw., vol. 38, no. 4, pp. 393–422,
Mar. 2002.
[6] I. Akyildiz, W. Su, Y. Sankarasubramaniam, and E. Cayirci, “A survey
on sensor networks,” IEEE Commun. Mag., vol. 40, no. 8, pp. 102–114,
Aug. 2002.
[7] A. Anastasopoulos, K. Bollas, D. Papasalouros, and D. Kourousis,
“Acoustic emission on-line inspection of rail wheels,” in Proc. 29th Eur.
Conf. Acoust. Emission Testing, Wien, Austria, 2010, pp. 1–8.
[8] E. Aw, “Low cost monitoring system to diagnose problematic rail bed:
case study of mud pumping site,” Ph.D. dissertation, Massachusetts Inst.
Technol., Cambridge, MA, USA, 2007.

[9] V. Balas and L. Jain, “World knowledge for sensors and estimators by
models and internal models,” J. Intell. Fuzzy Syst., vol. 21, no. 1, pp. 79–
88, Apr. 2010.
[10] D. Barke and W. Chiu, “Structural health monitoring in the railway
industry: A review,” Struct. Health Monitoring, vol. 4, no. 1, pp. 81–93,
Mar. 2005.
[11] P. Baronti et al., “Wireless sensor networks: A survey on the state of the
art and the 802.15. 4 and ZigBee standards,” Comput. Commun., vol. 30,
no. 7, pp. 1655–1695, May 2007.
[12] G. Barrenetxea, F. Ingelrest, G. Schaefer, and M. Vetterli, “The hitch-
hiker’s guide to successful wireless sensor network deployments,” in
Proc. 6th ACM Conf. Embedded Netw. Sens. Syst., Raleigh, NC, USA,
2008, pp. 43–56.
[13] S. Basagni, A. Carosi, and C. Petrioli, “Controlled vs. uncontrolled
mobility in wireless sensor networks: Some performance insights,” in
Proc. IEEE 66th Veh. Technol. Conf., Baltimore, MD, USA, 2007,
pp. 269–273.
[14] R. Baum et al., “Early warning of landslides for rail traffic between
Seattle and Everett,” in Proc. Landslide Risk Manage., Washington,
USA, 2005, pp. 731–740.
[15] C. Bell, “Event monitoring comes of age: ‘predict and prevent’ as well
as ‘find and fix’,” in Proc. IET Int. Conf. Railway Condition Monitor.,
Birmingham, U.K., 2006, pp. 44–47.
[16] V. Belotti, F. Crenna, R. Michelini, and G. Rossi, “Wheel-flat diagnostic
tool via wavelet transform,” Mech. Syst. Signal Process., vol. 20, no. 8,
pp. 1953–1966, Nov. 2006.
[17] P. Bennett et al., “Wireless sensor networks for underground railway
applications: Case studies in Prague and London,” Smart Struct. Syst.,
vol. 6, no. 5/6, pp. 619–639, 2010.
[18] E. Berlin and K. van Laerhoven, “Sensor networks for railway monitor-
ing: Detecting trains from their distributed vibration footprints,” in IEEE
Int. Conf. Distrib. Comput. Sens. Syst., Cambridge, MA, USA, 2013,
pp. 80–87.
[19] R. Bischoff, J. Meyer, O. Enochsson, G. Feltrin, and L. Elfgren, “Event-
based strain monitoring on a railway bridge with a wireless sensor
network,” in Proc. 4th Int. Conf. Struct. Health Monitor. Intell. Infras-
tructure, Zurich, Switzerland, 2009, pp. 1–8.
[20] M. Bocciolone, G. Bucca, A. Cigada, A. Collina, and L. Comolli, “An
application of FBG accelerometers for monitoring pantographs of under-
ground trains,” in Proc. SPIE 4th Eur. Workshop Opt. Fibre Sens., Porto,
Portugal, 2010, pp. 765341-1–765341-4.
[21] P. Boffi et al., “Optical fiber sensors to measure collector performance
in the pantograph-catenary interaction,” IEEE Sens. J., vol. 9, no. 6,
pp. 635–640, Jun. 2009.
[22] K. Bollas, P. Dimitrios, K. Dimitrios, and A. Athanasios, “Acoustic emis-
sion inspection of rail wheels,” J. Acoust. Emission, vol. 28, pp. 215–228,
Jan. 2010.
[23] B. Brickle et al., “Identification of Existing and New Technologies for
Wheelset Condition Monitoring,” TTCI Ltd., London, U.K., RSSB Rep.
Task T607, 2008.
[24] S. Bruni, R. Goodall, T. Mei, and H. Tsunashima, “Control and moni-
toring for railway vehicle dynamics,” Veh. Syst. Dyn., vol. 45, no. 7/8,
pp. 743–779, 2007.
[25] O. Bruno, A. Landi, M. Papi, and L. Sani, “Phototube sensor for moni-
toring the quality of current collection on overhead electrified railways,”
Proc. Inst. Mech. Eng. Part F: J. Rail Rapid Transit, vol. 215, no. 3,
pp. 231–241, Sep. 2001.
[26] K. Bruzelius, “An initial investigation on the potential applicability of
Acoustic Emission to rail track fault detection,” NDT E Int., vol. 37,
no. 7, pp. 507–516, Oct. 2004.
[27] S. Buggy et al., “Intelligent infrastructure for rail and tramways using
optical fibre sensors,” IET Railway Young Professionals’ Best Paper
Competition, 2011.
[28] B. Bukova and Z. Švecovà, “Smart dust in transport services,” Rail-
way Transp. Logist., vol. 3, no. 2, pp. 39–43, 2007.
[29] A. Cardenas et al., “Attacks against process control systems: Risk as-
sessment, detection, response,” in Proc. 6th ACM Symp. Inf., Comput.
Commun. Security, Hong Kong, 2011, pp. 355–366.
[30] A. Carrascal, A. Diez, and A. Azpeitia, “Unsupervised methods for ano-
malies detection through intelligent monitoring systems,” in Proc. 4th Int.
Conf. Hybrid Artif. Intell. Syst., Salamanca, Spain, 2009, pp. 137–144.
[31] V. Casola et al., “Freight train monitoring: A case-study for the
pSHIELD project,” in Proc. 6th Int. Conf. Innov. Mobile Internet Serv.
Ubiquitous Comput., Palermo, Italy, 2012, pp. 597–602.
[32] K. Chebrolu, B. Raman, N. Mishra, P. Valiveti, and R. Kumar, “Brimon:
A sensor network system for railway bridge monitoring,” in Proc. 6th Int.
Conf. Mobile Syst., Appl. Serv., Breckenridge, CO, USA, 2008, pp. 2–14.
[33] S. Cheekiralla, “Wireless sensor network-based tunnel monitoring,” in
Proc. 1st Workshop Real-World WSN Poster Session, 2005.
[34] H. Chen et al., “Towards scalable deployment of sensors and actuators
for industrial applications,” in IEEE Int. Conf. Sens. Netw., Ubiquitous
Trustworthy Comput., Taichung, Taiwan, 2008, pp. 285–290.
[35] J. Chen et al., “RAISE: RAIlway infrastructure health monitoring using
wireless sensor networks,” Sens. Syst. Softw., vol. 122, pp. 143–157,
2013.
[36] J. Chen and C. Roberts, “Effective condition monitoring of line side
assets,” in IET Int. Conf. Railway Condition Monitoring, Birmingham,
U.K., 2006, pp. 78–83.
[37] C.-Y. Chong and S. Kumar, “Sensor networks: Evolution, oppor-
tunities, challenges,” Proc. IEEE, vol. 91, no. 8, pp. 1247–1256,
Aug. 2003.
[38] S. Coraluppi and C. Carthel, “Multi-stage data fusion and the MSTWG
TNO datasets,” in Proc. 12th Int. Conf. Inf. Fusion, Seattle, WA, USA,
2009, pp. 1552–1559.
[39] R. da Costa Marques Pimentel et al., “Hybrid fiber-optic/electrical mea-
surement system for characterization of railway traffic and its effects
on a short span bridge,” IEEE Sens. J., vol. 8, no. 7, pp. 1243–1249,
Jul. 2008.
[40] N. Dadashi. et al., “A framework of data processing for decision mak-
ing in railway intelligent infrastructure,” in Proc. IEEE 1st Int. Multi-
Disciplinary Conf. Cognitive Methods Situation Awareness Decision
Support, Miami Beach, FL, USA, 2011, pp. 276–283.
[41] G. den Buurman and A. Zoeteman, “Quo Vadis: Review of the contribu-
tion of a weigh-in-motion system to optimising railway performance,”
in Proc. 7th World Congr. Railway Res., Montréal, QC, Canada, 2006,
pp. 1–10, [CD-ROM].
[42] R. Deuce and J. Rosinski, “Ultra miniature data loggers for testing and
in-service monitoring of railway axles and other rail vehicle compo-
nents,” in Proc. IET Int. Conf. Railway Condition Monitor., Birmingham,
U.K., 2006, pp. 57–60.
[43] K. Dierkx, “The Smarter Railroad. Transport R&D for Innovation
Project Report,” IBM Global Business Services, Armonk, NY, USA,
2009.
[44] M. Elia et al., “Condition monitoring of the railway line and overhead
equipment through onboard train measurement-an Italian experience,”
in Proc. IET Int. Conf. Railway Condition Monitor., Birmingham, U.K.,
2006, pp. 102–107.
[45] D. Estrin et al., “Connecting the physical world with pervasive net-
works,” IEEE Pervasive Comput., vol. 1, no. 1, pp. 59–69, Mar. 2002.
[46] G. Feltrin, “Wireless sensor networks: A monitoring tool for improving
remaining lifetime estimation,” in Civil Struct. Health Monitoring Work-
shop, Berlin, Germany, 2012, pp. 1–8.
[47] M. Filograno et al., “Real-time monitoring of railway traffic using
fiber Bragg grating sensors,” IEEE Sens. J., vol. 12, no. 1, pp. 85–92,
Jan. 2012.
[48] F. Flammini et al., “Towards wireless sensor networks for railway in-
frastructure monitoring,” in Proc. Electr. Syst. Aircraft, Railway Ship
Propulsion, Bologna, Italy, 2010, pp. 1–6.
[49] Y. Fukuta, “Possibility of sensor network applying for railway signal
system,” in Proc. 5th Int. Conf. Netw. Sens. Syst., Kanazawa, Japan,
2008, pp. 170–172.
[50] C. Gao et al., “Design of train ride quality testing system based on
wireless sensor network,” in Proc. Int. Conf. Electron. Mech. Eng. Inf.
Technol., Harbin, China, 2011, pp. 2636–2639.
[51] K. Ghosh et al., “Development of fishplate tampering detection system
for railway security based on wireless sensor network,” Int. J. Comput.
Appl., vol. 2, pp. 29–34, 2011.
[52] R. Goodall and C. Roberts, “Concepts and techniques for railway con-
dition monitoring,” in IET Int. Conf. Railway Condition Monitoring,
Birmingham, UK, 2006, pp. 90–95.
[53] C. Grosse et al., “Wireless acoustic emission sensor networks for struc-
tural health monitoring in civil engineering,” in Proc. Eur. Conf. Non-
Destructive Testing, Berlin, Germany, 2006, pp. 1–8.
[54] C. Grosse, S. Glaser, and M. Kruger, “Initial development of wireless
acoustic emission sensor Motes for civil infrastructure state monitoring,”
Smart Struct. Syst., vol. 6, no. 3, pp. 197–209, 2010.
[55] M. Grudén, A. Westman, J. Platbardis, P. Hallbjorner, and A. Rydberg,
“Reliability experiments for wireless sensor networks in train environ-
ment,” in Proc. Eur. Wireless Technol. Conf., 2009, pp. 37–40.
[56] D. Hall and J. Llinas, “An introduction to multisensor data fusion,” Proc.
IEEE, vol. 85, no. 1, pp. 6–23, Jan. 1997.
[57] J. Han and M. Kamber, “Data mining: Concepts and techniques,” in
Morgan Kaufmann Series in Data Management Systems. Amsterdam,
The Netherlands: Elsevier, 2006.

[58] T. Hay et al., “Transforming bridge monitoring from time-based
to predictive maintenance using acoustic emission MEMS sensors
and artificial intelligence,” in Proc. 7th World Congr. Railway Res.,
Montreal, Canada, 2006, CD-ROM.
[59] T. Hay, “Wireless Remote Structural Integrity Monitoring for Railway
Bridges,” Transportation Research Board, Washington, DC, DC, USA,
Tech. Rep. no. HSR-IDEA Project 54, 2007.
[60] M. Hendry, “The geomechanical behaviour of peat foundations below
rail-track structures,” Ph.D. dissertation, Univ. of Saskatchewan, Saska-
toon, SK, Canada, 2012.
[61] V. Hodge, Outlier and Anomaly Detection: A Survey of Outlier
and Anomaly Detection Methods. Saarbrucken, Germany: Academic,
2011.
[62] N. Hoult et al., “Wireless sensor networks: Creating ‘smart infrastruc-
ture’,” Proc. ICE-Civil Eng., vol. 162, no. 3, pp. 136–143, Aug. 2009.
[63] Y. Igarashi and S. Siomi, “Development of monitoring system for electric
switch machine,” Quarterly Rep. Railway Tech. Res. Inst., vol. 47, no. 2,
pp. 78–82, 2006, (Japan).
[64] K. Ishida, H. Kitabayashi, M. Nagasu, and K. Bekki, “New train control
and information services utilizing broadband networks,” Hitachi Rev.,
vol. 53, no. 1, pp. 21–24, 2004.
[65] A. Kerrouche et al., “Field tests of fibre Bragg grating sensors incorpo-
rated into CFRP for railway bridge strengthening condition monitoring,”
Sens. Actuators A, Phys., vol. 148, no. 1, pp. 68–74, Nov. 2008.
[66] J. Kim, K. S Lee, and J. Oh, “A study on the wireless onboard monitoring
system for railroad vehicle axle bearings using the SAW sensor,” Sens.
Syst. Softw., vol. 57, pp. 52–58, 2011.
[67] P. Kolakowski et al., “Structural health monitoring of a railway truss
bridge using vibration-based and ultrasonic methods,” Smart Mater.
Struct., vol. 20, no. 3, p. 035016, Mar. 2011.
[68] J. Konrad, S. Grenier, and P. Garnier, “Influence of repeated heavy axle
loading on peat bearing capacity,” in Proc. 60th Can. Geotech. Conf.,
Ottawa, ON, Canada, 2007, pp. 1551–1558.
[69] M. Kroener, “Energy harvesting technologies: Energy sources, genera-
tors and management for wireless autonomous applications,” in Proc.
9th Int. Multi-Conf. Syst., Signals Devices, Chemnitz, Germany, 2012,
pp. 1–4.
[70] M. Krüger et al., Sustainable Bridges. Technical Report on Wireless
Sensor Networks using MEMS for Acoustic Emission Analysis including
other Monitoring Tasks. Stuttgart, Germany: European Union, 2007.
[71] R. Lagnebäck, “Evaluation of wayside condition monitoring technolo-
gies for condition-based maintenance of railway vehicles,” M.S. thesis,
Luleâ Univ. of Technol., Luleâ, Sweden, 2007.
[72] C. Lai et al., “Development of a fiber-optic sensing system for train
vibration and train weight measurements in Hong Kong,” J. Sensors,
2012, Art. ID. 365165.
[73] A. Ledeczi et al., “Wireless acoustic emission sensor network for
structural monitoring,” IEEE Sens. J., vol. 9, no. 11, pp. 1370–1377,
Nov. 2009.
[74] C. Leung, S. Y. Zhang, S. W. Or, S. L. Ho, and K. Y. Lee, “Mag-
netoelectric smart current sensors for wireless condition monitoring of
train traction systems,” in Proc. 1st Int. Workshop High-Speed Intercity
Railways, Shenzhen, China, 2012, pp. 319–327.
[75] P. Li et al., “Estimation of railway vehicle suspension parameters for
condition monitoring,” Control Eng. Practice, vol. 15, no. 1, pp. 43–55,
Jan. 2007.
[76] R. Liu, Y. Wu, I. Wassell, and K. Soga, “Frequency diversity measure-
ments at 2.4 GHz for wireless sensor networks deployed in tunnels,” in
Proc. IEEE 20th Int. Symp. Pers., Indoor Mobile Radio Commun., 2009,
pp. 2990–2994.
[77] R. Liu, I. Wassell, and K. Soga, “Relay node placement for wireless
sensor networks deployed in tunnels,” in Proc. IEEE 6th Int. Conf.
Wireless Mobile Comput., Netw. Commun. , 2010, pp. 144–150.
[78] J. Lopez-Higuera, L. Rodriguez Cobo, A. Quintela Incera, and A. Cobo,
“Fiber optic sensors in structural health monitoring,” J. Lightw. Technol.,
vol. 29, no. 4, pp. 587–608, Feb. 2011.
[79] A. Lorestani, S. A. Mousavi, and R. Ebadaty, “Monitoring rail traffic us-
ing Wireless Sensor Network (WSN),” Int. J. Comput. Sci. Eng. Technol.,
vol. 2, no. 6, pp. 1280–1282, Jun. 2012.
[80] J. Lynch and K. Loh, “A summary review of wireless sensors and sensor
networks for structural health monitoring,” Shock Vibration Dig., vol. 38,
no. 2, pp. 91–130, Mar. 2006.
[81] S. Mahlknecht and S. Madani, “On architecture of low power wireless
sensor networks for container tracking and monitoring applications,” in
Proc. 5th IEEE Int. Conf. Ind. Inf., Vienna, Austria, 2007, pp. 353–358.
[82] T. Maly, M. Rumpler, H. Schweinzer, and A. Schoebel, “New devel-
opment of an overall train inspection system for increased operational
safety,” in Proc. IEEE Intell. Transp. Syst., Vienna, Austria, 2005,
pp. 188–193.
[83] F. Marquez, P. Weston, and C. Roberts, “Failure analysis and diagnostics
for railway trackside equipment,” Eng. Failure Anal., vol. 14, no. 8,
pp. 1411–1426, Dec. 2007.
[84] A. Matsumoto et al., “Continuous observation of wheel/rail contact
forces in curved track and theoretical considerations,” Veh. Syst. Dyn.,
vol. 50, pp. 349–364, 2012.
[85] S. McConnell and D. Skillicorn, “A distributed approach for prediction in
sensor networks,” in Proc. 1st Int. Workshop Data Mining Sens. Netw.a,
Newport Beach, CA, USA, 2005, pp. 28–37.
[86] M. McHutchon, W. J. Staszewski, and F. Schmid, “Signal processing for
remote condition monitoring of railway points,” Strain, vol. 41, no. 2,
pp. 71–85, May 2005.
[87] T. Mei and H. Li, “Measurement of vehicle ground speed using bogie-
based inertial sensors,” Proc. Inst. Mech. Eng. Part F, J. Rail Rapid
Transit, vol. 222, no. 2, pp. 107–116, Jun. 2008.
[88] L. Mottola et al., “Not all wireless sensor networks are created equal: A
comparative study on tunnels,” ACM Trans. Sens. Netw., vol. 7, no. 2,
Aug. 2010, Art. ID. 15.
[89] C. Nagode, M. Ahmadian, and S. Taheri, “Motion-based energy harvest-
ing devices for railroad applications,” in Proc. Am. Soc. Mech. Eng. Joint
Rail Conf., Urbana, IL, USA, 2010, pp. 267–271.
[90] A. Nair and C. Cai, “Acoustic emission monitoring of bridges: Re-
view and case studies,” Eng. Struct., vol. 32, no. 6, pp. 1704–1714,
Jun. 2010.
[91] W. Nan, M. Qingfeng, Z. Bin, L. Tong, and M. Qinghai, “Research on
linear wireless sensor networks used for online monitoring of rolling
bearing in freight train,” J. Phys., Conf. Ser., vol. 305, no. 1, p. 012024,
Jan. 2011.
[92] C. Nelson, S. Platt, D. Albrecht, V. Kamarajugadda, and M. Fateh,
“Power harvesting for railroad track health monitoring using piezoelec-
tric and inductive devices,” in Proc. SPIE Active Passive Smart Struct.
Integrated Syst., San Diego, CA, USA, 2008, vol. 6928, no. 1, pp. 1–9.
[93] B. Nejikovsky and E. Keller, “Wireless communications based system
to monitor performance of rail vehicles,” in Proc. IEEE/ASME Joint
Railroad Conf., Newark, NJ, USA, 2000, pp. 111–124.
[94] N. Nenov, E. N. Dirmitrov, G. S. Mihov, T. G. Ruzhekov, and
P. M. Piskulev, “A study on sensors for measuring load of railway vehicle
wheels in motion,” in Proc. 31st Int. Spring Seminar Electron. Technol.,
Budapest, Hungary, 2008, pp. 550–555.
[95] Art. ID. 012016R. Ngigi et al., “Modern techniques for condition moni-
toring of railway vehicle dynamics,” J. Phys., Conf. Ser., vol. 364, no. 1,
Jan. 2012.
[96] J. Nuffer and T. Bein, “Application of piezoelectric materials in
transportation industry,” in Proc. Global Symp. Innovative Solutions
Advancement Transp. Ind., San Sebastian, Spain, 2006, pp. 1–11.
[97] C. O’Donnell et al., “Pantograph damage and wear monitoring system,”
in Proc. IET Int. Conf. Railway Condition Monitoring, Birmingham,
U.K., 2006, pp. 178–181.
[98] B. Ollier, “Intelligent infrastructure—The business challenge: Railway
Condition Monitoring,” in Proc. IET Int. Conf. Railway Condition Mon-
itoring, Birmingham, U.K., 2006, pp. 1–6.
[99] M. Palo, “Condition monitoring of railway vehicles: a study on wheel
condition for heavy haul rolling stock,” M.S. thesis, Luleâ Univ. Tech-
nology, Luleâ, Sweden, 2012.
[100] A. Pascale, N. Varanese, G. Maier, and U. Spagnolini, “A wireless sensor
network architecture for railway signalling,” in Proc. 9th Italian Netw.
Workshop, Courmayeur, Italy, 2012, pp. 1–4.
[101] A. Pourghodrat, “Energy harvesting systems design for railroad safety,”
M.S thesis, Mechanical (and Materials) Engineering, Univ. Nebraska-
Lincoln, Lincoln, NE, USA, 2011.
[102] J. Rabatel, S. Bringay, and P. Poncelet, “SO_MAD: SensOr mining
for anomaly detection in railway data,” Adv. Data Mining Appl. Theor.
Aspects, LNCS, vol. 5633, pp. 191–205, 2009.
[103] J. Rabatel, S. Bringay, and P. Poncelet, “Anomaly detection in monitor-
ing sensor data for preventive maintenance,” Exp. Syst. Appl., vol. 38,
no. 6, pp. 7003–7015, Jun. 2011.
[104] J. Reason, H. Chen, R. Crepaldi, and S. Duri, “Intelligent Telemetry for
Freight Trains,” in Mobile Computing, Applications, Services, vol. 35.
Berlin, Germany: Springer-Verlag, 2010, pp. 72–91.
[105] J. Reason and R. Crepaldi, “Ambient intelligence for freight railroads,”
IBM J. Res. Develop., vol. 53, no. 3, pp. 1–14, May 2009.
[106] M. Reyer, S. Hurlebaus, J. Mander, and O. E. Ozbulut, “Design of a
wireless sensor network for structural health monitoring of bridges,”
in Proc. 5th Int. Conf. Sens. Technol., Palmerston North, New Zealand,
2011, pp. 515–520.

[107] D. Sala, J. Motylewski, and P. Koaakowsk, “Wireless transmission sys-
tem for a railway bridge subject to structural health monitoring,” Diag-
nostyka, vol. 50, no. 2, pp. 69–72, 2009.
[108] I. Salaberria, R. Carballedo, U. Gutierre, and A. Perallos, “Wireless
communications architecture for ‘train-to-earth’ communication in the
railway industry,” in Proc. 10th Int. Work-Conf. Artif. Neural Netw.,
Salamanca, Spain, 2009, pp. 626–633.
[109] G. Scholl et al., “SAW-based radio sensor systems for short-range appli-
cations,” IEEE Microw., vol. 4, no. 4, pp. 68–76, Dec. 2003.
[110] J. Schwarzenbach, L. Wilkinson, L. West, and M. Pilling, “Mapping
the remote condition monitoring architecture—Core report,” RSSB,
London, U.K., 2010.
[111] K. Sekula and P. Kolakowski, “Piezo-based weigh-in-motion system for
the railway transport,” Struct. Control Health Monitoring, vol. 19, no. 2,
pp. 199–215, Mar. 2012.
[112] G. Shafiullah, A. Gyasi-Agyei, and P. Wolfs, “Survey of wireless com-
munications applications in the railway industry,” in Proc. 2nd Int.
Conf. Wireless Broadband Ultra Wideband Commun., Sydney, NSW,
Australia, 2007, p. 65.
[113] G. Shafiullah, A. B. M. S. Ali, A. Thompson, and P. J. Wolfs, “Pre-
dicting vertical acceleration of railway wagons using regression algo-
rithms,” IEEE Trans. Intell. Transp. Syst., vol. 11, no. 2, pp. 290–299,
Jun. 2010.
[114] G. Shafiullah, S. A. Azad, and A. B. M. S. Ali, “Energy-efficient wireless
MAC protocols for railway monitoring applications,” IEEE Trans. Intell.
Transp. Syst., vol. 14, no. 2, pp. 649–659, Jun. 2013.
[115] J. Song, J. Zhong, and X. Li, “WHHT: A wireless sensor network for
heavy haul transportation,” in Proc. 6th Int. Conf. Wireless Commun.
Netw. Mobile Comput., Chengdu, China, 2010, pp. 1–4.
[116] F. Stajano et al., “Smart bridges, smart tunnels: Transforming wire-
less sensor networks from research prototypes into robust engi-
neering infrastructure,” Ad Hoc Netw., vol. 8, no. 8, pp. 872–888,
Nov. 2010.
[117] G. Stoneburner, A. Goguen, and A. Feringa, “Risk management
guide for information technology systems,” NIST Special Publication,
Gaithersburg, MD, USA, 2002, vol. 800, no. 30.
[118] TagMaster, “Gotcha—wheel flat detection and axle load measurement
system,” TagMaster North America Inc., Tacoma, WA, USA, Technical
Report, 2005.
[119] H. Tam et al., “Utilization of fiber optic Bragg Grating sensing systems
for health monitoring in railway applications,” Struct. Health Moni-
toring, Quantification, Validation, Implementation, vol. 1–2, pp. 1824–
1831, 2007.
[120] H. Tan, P. W. Q. Lee, W. K. G. Seah, and A. Zhi, “Impact of power
control in Wireless Sensor Networks Powered By Ambient Energy
Harvesting (WSN-HEAP) for railroad health monitoring,” in Proc.
Int. Conf. Adv. Inf. Netw. Appl. Workshops, Bradford, U.K., 2009,
pp. 804–809.
[121] N. Thakkar, J. A. Steel, and R. L. Reuben, “Rail-wheel interaction mon-
itoring using Acoustic Emission: A laboratory study of normal rolling
signals with natural rail defects,” Mech. Syst. Signal Process., vol. 24,
no. 1, pp. 256–266, Jan. 2010.
[122] S. Tilak et al., “A taxonomy of wireless micro-sensor network models,”
ACM SIGMOBILE Mobile Comput. Commun. Rev., vol. 6, no. 2, pp. 28–
36, Apr. 2002.
[123] C. Townsend and S. Arms, “Wireless Sensor Networks. Principles and
Applications,” in Sensor Technology Handbook, J. Wilson, Ed. Oxford,
U.K.: Elsevier, 2005, ch. 22.
[124] H. Tsunashima, T. Kojima, Y. Marumo, H. Matsumoto, and T. Mizuma,
“Condition monitoring of railway track using in-service vehicle,” in
Proc. 4th IET Int. Conf. Railway Condition Monitoring, 2008, pp. 1–6.
[125] H. Waki, “Monitoring of contact line equipment by trains in operation,”
JR East Tech. Rev., East Japan Railway Culture Found., vol. 17, pp. 11–
14, 2010.
[126] C. Ward et al., “Condition monitoring opportunities using vehicle-based
sensors,” Proc. Inst. Mech. Eng. Part F, J. of Rail Rapid Transit, vol. 225,
no. 2, pp. 202–218, 2011.
[127] C. Wei et al., “Real-time train wheel condition monitoring by fiber Bragg
grating sensors,” Int. J. Distrib. Sens. Netw., vol. 2012, pp. 1–7, 2012,
Art. ID. 409048.
[128] A. Wilkinson, “Long range inspection and condition monitoring of rails
using guided waves,” presented at the Proc. 12th Int. Conf. Exhib.,
Railway Eng., London, U.K., 2013.
[129] R. Willsch, W. Ecke, and G. Schwotzer, “Spectrally encoded optical fibre
sensor systems and their application in process control, environmental
and structural monitoring,” in Proc. SPIE Opt. Fibers, Appl., 2005,
vol. 5952, doi:10.1117/12.619430.
[130] M. Wischke, M. Masur, M. Kroner, and P. Woias, “Vibration harvesting
in traffic tunnels to power wireless sensor nodes,” Smart Mater. Struct.,
vol. 20, no. 8, p. 085014, Aug. 2011.
[131] P. Wolfs, S. Bleakley, S. T. Senini, and P. Thomas, “An autono-
mous, low cost, distributed method for observing vehicle track interac-
tions,” in Proc. IEEE/ASME Joint Rail Conf., Atlanta, GA, USA, 2006,
pp. 279–286.
[132] R. Wong, R. Thomson, and E. Choi, “In situ pore pressure responses
of native peat and soil under train load: A case study,” J. Geotech.
Geoenviromental Eng., vol. 132, no. 10, pp. 1360–1369, Oct. 2006.
[133] H. Yazdi., “Intelligent condition monitoring of railway signalling
equipment using simulation,” in Proc. Inst. Elect. Eng.—Seminar
Condition Monitoring Rail Transport Syst. (Ref. No. 1998/501), 1998,
pp. 13-1–13-5.
[134] J. Yick, B. Mukherjee, and D. Ghosal, “Wireless sensor network survey,”
Comput. Netw., vol. 52, no. 12, pp. 2292–2330, Aug. 2008.
[135] P. Yilmazer et al., “The structural health condition monitoring of rail
steel using acoustic emission techniques,” in Proc. 51st Annu. Conf.
NDT, Daventry, U.K., 2012, pp. 1–12.
[136] F. Zhou, M. D. Duta, M. P. Henry, S. Baker, and C. Burton, “Remote
condition monitoring for railway point machine,” in Proc. ASME/IEEE
Joint Railroad Conf., Washingtonm, DC, DC, USA, 2002, pp. 103–108.
[137] H. Zimmermann, “OSI reference model-the ISO model of architecture
for open systems interconnection,” IEEE Trans Commun., vol. 28, no. 4,
pp. 425–432, Apr. 1980.
[138] S. Zöller et al., “Efficient real-time monitoring of multimodal transports
with wireless sensor networks,” presented at the Proc. 36th IEEE Conf.
Local Comput. Netw., Bonn, Germany, 2011.
Victoria J. Hodge received the B.Sc. (Hons) and
Ph.D. degrees in computer science from Univer-
sity of York, York, U.K., in 1997 and 2001,
respectively.
Since 2001 she has been a Research Associate
with the Department of Computer Science, Uni-
versity of York. She is currently developing an
infrastructure for condition monitoring of railway
networks. She is on the editorial board of two jour-
nals and has published two books and two book
chapters as well as 40 articles and proceedings. Her
research interests include outlier detection, condition monitoring, neural net-
works, pattern recognition, data mining, and information retrieval.
Dr. Hodge was a member of the research group that won the Outstanding
Engineering Research Team of the Year award at the Times Higher Education
Awards 2011.
Simon O’Keefe received the M.A. degree in en-
gineering and management from University of
Cambridge, Cambridge, U.K., in 1986; the M.Sc.
degree in operational research from Lancaster Uni-
versity, Lancaster, U.K., in 1987; and the M.Sc. de-
gree in information processing and the D.Phil. degree
in computer science from University of York, York,
U.K., in 1993 and 1997, respectively.
He is currently a Lecturer with the Department
of Computer Science, University of York. He was
previously an Operational Research Analyst with
Her Majesty’s Customs and Excise, London, U.K. He is a member of the
York Centre for Complex Systems Analysis. His research interests principally
include neural networks and pattern recognition, data mining, and novel com-
putation.
Dr. O’Keefe is a Fellow of the Higher Education Academy.

Michael Weeks received the B.Eng. (Hons) de-
gree in computer engineering and the M.Phil. de-
gree in electronics from University of Teesside,
Middlesbrough, U.K, in 1997 and 2006, respectively.
Since 2000 he has been a Research Associate with
the Department of Computer Science, University
of York, York, U.K. working on several research
projects related to neural networks. He is currently
working on two projects; a web-based infrastruc-
ture for condition monitoring across sensor networks
and an internationally recognized web-based virtual
laboratory for neuroscientists. Between 1997 and 2000 he was a Research
Assistant with University of Teesside in the area of media server architectures.
Mr. Weeks received the Sony Prize for Electronics while studying at Univer-
sity of Teesside. He was also a member of the research group at York that won
the Outstanding Engineering Research Team of the Year award at the Times
Higher Education Awards 2011.
Anthony Moulds received the B.Sc. (Hons) degree
in electronics engineering from University of Kent,
Kent, U.K., in 1984 and the M.Sc. degree in informa-
tion technology from University of York, York, U.K.,
in 1990.
After graduating in 1984, he spent four years as
an Electronics Design Engineer with a subsidiary
company of Westland Helicopters, U.K., and later
was with the Department of Computer Science, Uni-
versity of York, in 1990 as an Experimental Officer.
He is currently a Senior Experimental Officer within
the Department. He has particular expertise in the design and development of
novel research hardware with a focus on high-speed circuit design and field-
programmable gate array implementation.
Mr. Moulds is a member of the Institution of Engineering and Tech-
nology, U.K.

Wireless sensor networks for condition monitoring in the railway industry a survey

More Related Content

What's hot (20)

Viewers also liked (20)

Similar to Wireless sensor networks for condition monitoring in the railway industry a survey (20)

More from redpel dot com (20)

Recently uploaded (20)

Wireless sensor networks for condition monitoring in the railway industry a survey